AlphaGenome 2: source-linked RNA and regulatory findings
Research update · 14 November 2026
14 research contexts connect existing observations with AlphaGenome 2 (AG2) predictions. They cover human SMN1, SMN2, PLS3, MDM4 and PERP/ARFGEF3, plus MAPK12 RNA paths, TP73 first-junction coverage, STMN2 cryptic-site representation and a distinct mouse Hspa8 question. Each context preserves the source, exact identity, model output and strongest counterfinding. These are different questions and reused sources; the count is not a count of independent biological confirmations.
Model channels labelled observed refer to channel provenance. The predicted variant response is not a new experimental measurement. Model scores, attribution and splice-support fractions do not establish binding, causality or treatment efficacy.
AlphaGenome 2 (AG2): SMN2 allele and expression-model context
A13 · A43 · Source-linked research context
Existing natural-allele cohort associations and AlphaGenome 2 GeneMaskLFCScorer outputs are shown as distinct evidence. The model reports small DIFF_MEAN values across a shared chr5 interval; these are predictions, not measured RNA changes.
Source and interpretation
The Ruhno allele set provides allele identities and cohort associations, not isolated allele interventions. In the model output, observed describes channel provenance; it does not mean the variant effect was experimentally observed. A gene-first follow-up asks which sequence contributes to the expression objective; attribution is not the fraction of expression explained.
Model and output provenance
These are API predictions made with the AlphaGenome 2 client. ALL_FOLDS is the requested model selector; an immutable served weight revision was not independently supplied. Channel labels such as observed or imputed describe output provenance, not an experiment on the variant. Model outputs and source measurements retain their own units and denominators.
GeneMaskLFCScorer DIFF_MEAN, common interval GRCh38 chr5:[69551712,70600288) · 4 rows
Allele
C1 adult tissue UBERON:0006469
In-vitro motor neuron CL:0000100
Channel status
A-549G; GRCh38 chr5:70075972 A>G
0.0017555952
Not reported
motor-neuron value not recoverably saved
A-44G; GRCh38 chr5:70076477 A>G
0.0029186606
0.0025262833
C1 observed; motor-neuron imputed_supported
C-1897T; GRCh38 chr5:70074624 C>T
0.00044804811
0.0006043911
C1 observed; motor-neuron imputed_supported
C-478T; GRCh38 chr5:70076043 C>T
-0.0001321435
-0.00032401085
C1 observed; motor-neuron imputed_supported
Interpretation limits
A44G has an AG1/AG2 sign disagreement; the AG1 serving equation and a common effect scale are unresolved.
A549G in CL:0000100 is missing and was not recovered; values for that channel in other alleles are imputed_supported.
C1 is heterogeneous adult tissue, not isolated SMA motor neurons. Phase, causal effect, training overlap, and gene-mask annotation equivalence remain unresolved.
Gao HEK293 shortened-minigene PSI measurements differ by substitution: C>A increased inclusion while C>G reduced it. Native-genome AG2 outputs have mixed signal and fraction behavior and do not reproduce a measured PSI endpoint. Separate c.840T>C and c.859G>C positive-source comparisons and A-44G are included with their counterfindings.
Comparison boundary
The source assay is a HEK293 minigene. The AG2 prediction uses native genomic context and predicted junction outputs. Cell, construct, and endpoint differ, so the values are not a direct validation comparison. A source-matched measured endpoint is the next useful evidence.
Model and output provenance
These are API predictions made with the AlphaGenome 2 client. ALL_FOLDS is the requested model selector; an immutable served weight revision was not independently supplied. Channel labels such as observed or imputed describe output provenance, not an experiment on the variant. Model outputs and source measurements retain their own units and denominators.
Additional source-positive comparisons
The same C1 spinal-cord configuration gives the expected inclusion-support direction for c.840T>C, but the opposite direction for source-positive c.859G>C. A-44G has no matched isolated-allele negative-control assay here. This does not establish calibrated splice-effect prediction. C1 spinal tissue and the HEK293 minigene are different contexts.
The native substitutions are not authenticated as identical to the complete shortened minigene constructs.
For both substitutions, donor and acceptor fractions increase while raw junction changes are mixed; competing-junction decline, especially skip J68, drives the fraction arithmetic.
These model outputs are not measured PSI, rescue, or causal mechanism evidence; the diagnostic decomposition is post hoc on the same responses.
AlphaGenome 2 (AG2): PLS3 rs73226811: eQTL and RNA-model context
A44 · Source-linked research context
The exact GRCh38 chrX:116225577 C>A allele is rs73226811. A GTEx v8 Testis conditional eQTL reports NES +0.0769585; AG2 DIFF_MEAN is negative in Testis and slightly positive in an imputed_supported motor-neuron channel.
Identity and context
PLS3-AS1 is a distinct gene and is not combined with PLS3. The gene-centered model window does not attribute this distal allele; an exact linked distal source or a separately documented mechanism would be needed.
Model and output provenance
These are API predictions made with the AlphaGenome 2 client. ALL_FOLDS is the requested model selector; an immutable served weight revision was not independently supplied. Channel labels such as observed or imputed describe output provenance, not an experiment on the variant. Model outputs and source measurements retain their own units and denominators.
Exact allele source and model values · 3 rows
Context
Value
Interpretation
GTEx v8 Testis conditional eQTL
NES +0.0769585
Association; rank 2; tssDistance 664403
AG2 Testis UBERON:0000473
DIFF_MEAN -0.00096169114
Observed channel provenance
AG2 in-vitro motor neuron CL:0000100
DIFF_MEAN +0.00008480996
Imputed_supported channel
Interpretation limits
The eQTL is an association that may reflect linkage disequilibrium or conditioning, not an isolated-allele causal effect.
Fine-mapping returned zero records in the queried endpoint; this does not establish that fine-mapping evidence is absent.
Five assay labels are correlated outputs, not independent replications; possible GTEx training overlap remains unresolved. No motor-neuron effect or SMA protection is established.
Two reference RNA-expression attribution maps identify a shared leading TSS/first-exon tile. The terminal-exon tile ranks second in C1 spinal tissue and third in the in-vitro motor-neuron context; an internal-exon tile ranks second in the latter. These are sequence-attribution maps, not measured regulatory effects.
Next source question
Separate exonic from non-exonic attribution in the two specified follow-up regions and reuse existing CAGE, PRO-cap, and accessible-chromatin evidence by coordinate before considering any additional inference.
Model and output provenance
These are API predictions made with the AlphaGenome 2 client. ALL_FOLDS is the requested model selector; an immutable served weight revision was not independently supplied. Channel labels such as observed or imputed describe output provenance, not an experiment on the variant. Model outputs and source measurements retain their own units and denominators.
GRCh38 half-open 500-bp attribution regions · 4 rows
Context
Rank
Region
Annotated overlap
Both objectives
1
chr5:[70049500,70050000)
TSS and 98 bp of first exon
C1 spinal tissue
2
chr5:[70077000,70077500)
482 bp of terminal exon
In-vitro motor neurons
3
chr5:[70077000,70077500)
482 bp of terminal exon
In-vitro motor neurons
2
chr5:[70067000,70067500)
287 bp of internal exons
Interpretation limits
High attribution overlaps RNA objective exons and does not identify an enhancer.
Attribution shares are not fractions of expression explained; no null or functional perturbation is available.
AlphaGenome 2 (AG2): SMN2 regulatory-coordinate and chromatin context
A43 · Source-linked research context
Three fixed regulatory intervals were cross-checked against six distinct donor ATAC experiments, comprising three healthy and three ALS donors. No selected conservative peak overlaps were found; this is a bounded source-selection result, not evidence of biological inactivity.
Interpretation
The coordinate overlap check does not test whether the regions regulate SMN2 in relevant cells. It only reports the selected interval and peak comparison.
Model and output provenance
These are API predictions made with the AlphaGenome 2 client. ALL_FOLDS is the requested model selector; an immutable served weight revision was not independently supplied. Channel labels such as observed or imputed describe output provenance, not an experiment on the variant. Model outputs and source measurements retain their own units and denominators.
Fixed intervals and selected ATAC comparison · 1 rows
Source set
Donors
Selected conservative peak overlap
Scope
Six donor ATAC experiments
3 healthy; 3 ALS
0 at all three fixed intervals
No SMA donors; not an inactivity result
Interpretation limits
The experiments contain no SMA donors and do not establish SMN2 target-gene causality.
A shared exclusion mask overlaps none of the three intervals, so this specific mask does not explain the zero selected overlaps; broader activity and other sources remain untested.
AlphaGenome 2 (AG2): MDM4 reference splice topology and sample-source boundary
A13 · Source-linked research context
Reference splice-event summaries, local junction edges, strict read paths, and six named BAM paths describe different evidence units. Their coordinate topology can be compared, but they do not establish a selective MDM4 isoform switch in the SMA cell line.
Counterfinding
The same named edges can have event-specific counts. SE inclusion counts are not the sum of two edge columns, and A3SS inclusion evidence spans a short boundary. Strict read-name groups are separately filtered local assignments, not validated phased transcripts.
Model and output provenance
These are API predictions made with the AlphaGenome 2 client. ALL_FOLDS is the requested model selector; an immutable served weight revision was not independently supplied. Channel labels such as observed or imputed describe output provenance, not an experiment on the variant. Model outputs and source measurements retain their own units and denominators.
Relationship to the existing human RNA observations
The source cohort contains 15 deposited libraries from one GM03813 fibroblast line. Local 68-nt, 39-nt and skip paths are separate alignment evidence. A processing log naming six BAM paths does not authenticate their input contents or join AG2 reference edges to sample-level treatment effects. Original event membership and denominators remain unresolved.
Distinct evidence units retained · 4 rows
Evidence unit
What it supports
What it does not establish
Reference splice event
Event-level inclusion or choice
Local read paths or isoform fractions
Reference junction edge
Coordinate topology
Phased transcript chain
Strict joined read path
Filtered local assignment
Historical run membership or sample effect
Six BAM path names
Names recorded in processing log
Verified sample-level quantitative join
Interpretation limits
Generic primary dermal fibroblasts and static reference DNA do not match GM03813 SMA or treatment.
Individual edges are not phased chains. SE event counts and A3SS choices cannot be converted directly into three isoform fractions.
The exact run build, processing options, input hashes, and event membership for the six BAM paths remain unbound; do not join the separately named paths to sample effects.
AlphaGenome 2 (AG2): PLS3 reference annotation and gene-first model context
A44 · Source-linked research context
PLS3 reference annotations differ by Ensembl release. A gene-first splice scorer returned four alternative edges as not_returned and a positive canonical edge score; neither result establishes absence of alternative splicing or a measured expression change.
Existing RNA evidence
Four verified transcript-quantification tables from one ENCODE motor-neuron experiment include PLS3.17 and canonical-stable ENST00000355899.7. Exact coordinate-bound junction counts are the next relevant measured source; keep assay and phasing explicit.
Model and output provenance
These are API predictions made with the AlphaGenome 2 client. ALL_FOLDS is the requested model selector; an immutable served weight revision was not independently supplied. Channel labels such as observed or imputed describe output provenance, not an experiment on the variant. Model outputs and source measurements retain their own units and denominators.
Distinct PLS3 reference regions — GRCh38 half-open · 2 rows
Context
Region
Selection rank
Boundary
C1 spinal tissue; GTEx polyA RNA
chrX:[115603000,115603500)
1
500 bp; 90 bp exon overlap in release 116
In-vitro motor neuron; ENCODE total RNA
chrX:[115563500,115564000)
2
500 bp; rank 1 lies in excluded control window;139 bp exon overlap in release116
Reference annotation and model output · 3 rows
Measure
Release/result
Interpretation
Exonic overlap
Release 112: 0 bp in both selected regions; release 116: 90 bp in C1 and 139 bp in motor neurons
Reference annotation only
Alternative splice edges
4 not_returned
Not an absence or zero result
Canonical edge
1.958001852
Model score; not measured PSI
Interpretation limits
Release 112 and release 116 annotations differ; exonic overlap cannot be treated as a biological change.
not_returned is not a zero score or absence claim. No exact junction counts were supplied in the inspected experiment files.
Transcript quantification tables represent two biological replicate labels and two output methods, not four independent samples; newer alternative stable IDs are absent from the target catalogs.
AlphaGenome 2 (AG2): PERP and ARFGEF3: neuronal regulatory-source context
A21 · Source-linked research context
The fixed PERP-linked element EH38E2507697 has a verified neuronal assay connection, but target-gene causality remains open. A nearby shared 450 bp element has multiple target IDs, including PERP and ARFGEF3; spatial overlap and model attribution do not establish exclusive regulation.
Strongest counterfinding
At EH38E2507697, the reported neuronal source values are ATAC z=1.34, H3K4me3 z=-0.12, H3K27ac z=-0.56, with CTCF missing. The source establishes a neuronal measurement connection, not an active enhancer or PERP functional link. The adjacent ARFGEF3 evidence also remains non-specific.
Next source question
Identify exact biosamples and files for the element and obtain measured neuronal PERP transcript or 3D linkage with target identity. Preserve the PERP.12 versus current .13 annotation distinction.
Model and output provenance
These are API predictions made with the AlphaGenome 2 client. ALL_FOLDS is the requested model selector; an immutable served weight revision was not independently supplied. Channel labels such as observed or imputed describe output provenance, not an experiment on the variant. Model outputs and source measurements retain their own units and denominators.
A separate 450-bp source-selected element
At GRCh38 chr6:[138115850,138116300), the EPIraction source contains eight target-gene IDs including PERP and ARFGEF3. This is a different region from EH38E2507697. A separately requested ARFGEF3 RNA attribution map has about 2.98-fold mean absolute attribution versus each immediate same-width flank; the element contributes 0.0978818% of the absolute attribution in that returned window. Different gene-centered windows prevent a cross-gene effect comparison. An exact allele or target-resolved functional observation is needed to distinguish target hypotheses.
Context comparability
The additional dermal-fibroblast RNA attribution used a minus-strand track, whereas the motor-neuron RNA track was unstranded. Their numerical difference is not an isolated cell-type effect. These DNA/RNA regulatory comparisons do not test the separate PERP–DAG1 protein-interaction hypothesis.
Bounded source and model findings · 4 rows
Finding
Result
Boundary
SCREEN neuronal element
ATAC z 1.34; H3K4me3 z -0.12; H3K27ac z -0.56; CTCF missing
Ancillary collection; no target-link result
ATAC cCRE/flank ratio
5.686308134
Model consistency, not independent validation
H3K27ac cCRE/flank ratio
0.867984887
No local elevation in this comparison
Shared element
450 bp; eight target IDs
No exclusive PERP or ARFGEF3 assignment
Separate source-selected ARFGEF3-map question · 1 rows
Fixed region, GRCh38 half-open
Local mean-absolute-attribution ratios
Own-window absolute-attribution share
Source target identities
chr6:[138115850,138116300)
2.98160 versus left flank;2.98040 versus right flank
0.0978818%
8 target-gene IDs; not an exclusive PERP enhancer
Interpretation limits
Human SMA/control PERP ranges overlap; donor mapping and processing remain unresolved. Separate NGN2 comparison is adjusted P=.929.
The motor-neuron assay source includes one ALS donor; iNPC/iPSC-derived motor-neuron biosamples differ. CTCF is missing and no displayed target-link result confirms PERP causality.
No selected thresholded peak overlaps the original cCRE/tile/fixed ±5 kb regions; this is not evidence of zero activity. The shared element is not target-specific.
For the mouse Hspa8 G470R allele, one paired broad-brain prediction returned small raw-score ratios for fixed upstream and downstream junctions. A separate diagnostic skip ratio is larger but uses a low reference score. This does not establish a cis-RNA effect.
Context separation
The source mouse allele is mm10 chr9:40804249 G>C, p.Gly470Arg. NM_031165.5:c.1408G>C is our current annotation crosswalk, not an original-paper transcript version. The prediction uses broad mouse brain UBERON:0000955, ENCODE polyA RNA; this is not an SMA motor-neuron or developmental-stage model. The known protein-level and SMN2-trans observations are separate from this local Hspa8 cis-RNA question.
Model and output provenance
These are API predictions made with the AlphaGenome 2 client. ALL_FOLDS is the requested model selector; an immutable served weight revision was not independently supplied. Channel labels such as observed or imputed describe output provenance, not an experiment on the variant. Model outputs and source measurements retain their own units and denominators.
AlphaGenome 2 (AG2): Reported PLS3 c.367+182C>T blood association
A44 · Source-linked research context
A reported association links PLS3 c.367+182C>T to blood expression in a 50-person SMA and 21-control cohort. The genomic coordinate, transcript identity, carrier count, and original supplement were not established, so no AG2 variant call was made.
Next source question
Recover the original supplement and bind the exact transcript and genomic allele before considering any variant-level model comparison. Keep this candidate separate from rs73226811.
Reported association and strongest counterfinding
The source reports p=0.04 and a carrier median of 19.1 copies/ng cDNA. The pooled 10.8 value is inconsistently labelled mean or median and is not an authenticated noncarrier comparator. Carrier count is not bound. Family severity directions differ, and higher expression does not establish consistent clinical benefit.
Reported cohort context · 4 rows
Item
Reported information
Status
Cohort
50 SMA; 21 controls
Source report
Variant
PLS3 c.367+182C>T
Genomic identity pending
Carrier count
Not bound
Missing
AG2 result
None
No call made
Interpretation limits
The exact variant identity is unresolved; no genomic coordinate or transcript is inferred.
The reported blood association has no bound carrier count and no consistent clinical benefit.
Venous-blood reference RNA differs from full-length ddPCR; a literal blood RNA label was imputed. This is not a motor-neuron result.
AlphaGenome 2 (AG2): SMN1 c.863G>T splice-source and model context
A13 · Source-linked research context
SMN1 NM_000344.3 c.863G>T maps to GRCh38 chr5:70951969 G>T. The shortened HEK293 minigene showed exon 7 inclusion decreasing from 91.7% to 14.8%; native blood and fibroblast observations are separate materials. AG2 fraction changes are tiny and negative while raw J78 rises in all three imputed channels.
Bounded interpretation
This is a source/model comparison, not reproduction of the strong experimental splice effect. The finite reference-transfer question is closed; the useful next evidence would be source-bound human native-RNA allele/junction data with material and denominator specified.
Model and output provenance
These are API predictions made with the AlphaGenome 2 client. ALL_FOLDS is the requested model selector; an immutable served weight revision was not independently supplied. Channel labels such as observed or imputed describe output provenance, not an experiment on the variant. Model outputs and source measurements retain their own units and denominators.
Source and model comparison · 4 rows
Evidence
Result
Context
HEK293 shortened minigene
Exon 7 inclusion 91.7% to 14.8%
Experimental construct
AG2 inclusion-support fractions
Negative changes across 3 channels; 0.0055–0.0173 percentage points
Coupled imputed channels
AG2 raw J78
Increases in all 3 channels
Raw junction score, not PSI
Native RNA
Blood from 2 patients; fibroblast from 1 patient
Distinct source materials; no matched denominator
HEK293 imputed model channels — ALT-minus-REF fractions expressed in percentage points · 3 rows
The model fraction changes are only 0.0055–0.0173 percentage points and are not PSI; rising raw J78 is compatible with falling J78 fraction because the denominator also rises.
Native RNA observations are not a matched model denominator and no quantitative conversion is established.
SMN1 and SMN2 remain distinct paralogs; do not infer SMN2 behavior from this result.
Five reference-model junctions are linked by exact coordinates to existing human fibroblast RNA outputs. The sequential exon 14-to-15-to-16 partial chain is present in stored alignment patterns in all six libraries; a strictly exclusive two-isoform interpretation is not established.
Exact identity and source material
Human MAPK12 (p38-gamma), ENSG00000188130.14, GRCh38 chromosome 22 negative strand. MAPK11/p38-beta and MAPK14/p38-alpha remain distinct. Public study SRP334251 reports GM03813 fibroblasts with 24-hour risdiplam 1000 nM (HiR: SRR15622462/61/60) and shared DMSO (SRR15622470/69/63). Three libraries per arm do not establish three independent donors or authenticated culture units.
Reference model and output provenance
One source-defined reference interval chr22:[49724894,50773470) was queried for SPLICE_JUNCTIONS in generic primary dermal fibroblasts UBERON:0002097_CL:0002551, with ALL_FOLDS requested. The exact served weight revision was not independently supplied. All three actual response channels are retained below; the primary ENCODE channel was chosen from saved metadata before inference.
What the native source adds
The four MXE component arrays were already stored in the earlier source analysis. Their negative-strand coordinate mapping and the central SE 24567 edge are now linked to the five model boundaries. Ordered source-derived patterns join the central and adjacent junctions within the same partial 14-to-15-to-16 alignment pattern:17/17/16 HiR and13/19/26 DMSO. The short 13-to-15-to-16 alternative contributes two aggregate alignment units in SRR15622469; these are not two independent molecules.
Counting rules and strongest counterfinding
For the fifth skipping-group library, component counts sum to 19 while the event aggregate is 17. The documented counting semantics allow a two-junction contribution to enter both component columns and the event aggregate once. The two-unit residual is also present in the stored pattern keys. Do not sum reused event counts. Source JC deltaPSI 0.203 / FDR 0.01656 versus JCEC 0.148 / FDR 0.16317 remains a counting-mode-sensitive observation; the central SE event has FDR 1.0.
Next discriminating question
Which existing long-range RNA or linked-fragment evidence can distinguish complete MAPK12 transcript structures beyond the observed partial chain? Neither the reference model nor these 100-nt products establish a protein isoform switch or explain the JC/JCEC difference. A full-chain claim requires a genuinely informative source; the same reference or short-pattern computation should not be repeated.
AG2 raw junction signals: model channels remain separate · 15 rows
Junction
Actual returned track
Channel provenance
Raw model value
right_to_alt1
UBERON:0002097_CL:0002551_polyA plus RNA-seq-encode
imputed_exploratory
0.0343529917299747
alt1_to_left
UBERON:0002097_CL:0002551_polyA plus RNA-seq-encode
imputed_exploratory
20.63564682006836
right_to_alt2
UBERON:0002097_CL:0002551_polyA plus RNA-seq-encode
imputed_exploratory
24.805028915405277
alt2_to_left
UBERON:0002097_CL:0002551_polyA plus RNA-seq-encode
imputed_exploratory
1.5381826162338257
sequential_alt2_to_alt1
UBERON:0002097_CL:0002551_polyA plus RNA-seq-encode
imputed_exploratory
18.892269134521484
right_to_alt1
UBERON:0002097_CL:0002551_polyA plus RNA-seq-gtex
imputed_exploratory
0.0199042055755853
alt1_to_left
UBERON:0002097_CL:0002551_polyA plus RNA-seq-gtex
imputed_exploratory
19.26856803894043
right_to_alt2
UBERON:0002097_CL:0002551_polyA plus RNA-seq-gtex
imputed_exploratory
23.071720123291016
alt2_to_left
UBERON:0002097_CL:0002551_polyA plus RNA-seq-gtex
imputed_exploratory
0.9553375840187072
sequential_alt2_to_alt1
UBERON:0002097_CL:0002551_polyA plus RNA-seq-gtex
imputed_exploratory
16.90708351135254
right_to_alt1
UBERON:0002097_CL:0002551_total RNA-seq-encode
observed
0.0134973041713237
alt1_to_left
UBERON:0002097_CL:0002551_total RNA-seq-encode
observed
15.806010246276855
right_to_alt2
UBERON:0002097_CL:0002551_total RNA-seq-encode
observed
19.22203826904297
alt2_to_left
UBERON:0002097_CL:0002551_total RNA-seq-encode
observed
0.3781187236309051
sequential_alt2_to_alt1
UBERON:0002097_CL:0002551_total RNA-seq-encode
observed
13.338152885437012
Existing event-conditioned junction counts; three libraries per source arm · 5 rows
Junction
GRCh38 intron [start,end), strand
HiR library counts
DMSO library counts
Source count field
right_to_alt1
chr22:[50248890, 50249634] (-)
0,0,0
0,2,1
MXE2794 printed second_to_downstream_count
alt1_to_left
chr22:[50248751, 50248830] (-)
21,14,31
32,17,35
MXE2794 printed upstream_to_second_count
right_to_alt2
chr22:[50249454, 50249634] (-)
87,70,55
48,24,40
MXE2794 printed first_to_downstream_count
alt2_to_left
chr22:[50248751, 50249102] (-)
1,2,2
3,0,0
MXE2794 printed upstream_to_first_count
sequential_alt2_to_alt1
chr22:[50248890, 50249102] (-)
37,32,33
26,35,51
SE24567 upstream_to_target_count
Existing100-nt ordered alignment-pattern contributions; not molecule or donor counts · 6 rows
Public library
Source condition
Read length (nt)
Partial14-to15-to16 chain
Short13-to15-to16 path
SRR15622462
HiR1000 treatment
100
17
0
SRR15622461
HiR1000 treatment
100
17
0
SRR15622460
HiR1000 treatment
100
16
0
SRR15622470
shared DMSO
100
13
0
SRR15622469
shared DMSO
100
19
2
SRR15622463
shared DMSO
100
26
0
Interpretation limits
AG2 predicts generic dermal-fibroblast reference junctions, not GM03813, risdiplam exposure, TDP43 state or SMA response. Observed labels describe channel provenance; all returned values remain predictions.
Native event-conditioned counts and model raw signals are different units. Do not calculate model/count ratios, PSI, treatment effects or a common efficacy scale.
The 100-nt pattern products contain aggregated alignment contributions without QNAME, mate linkage or complete molecules. They do not establish full four-exon transcripts, protein products or kinase activity.
The central SE event has FDR 1.0. The MXE JC and JCEC contrasts reuse the same libraries and differ in qualification; source genotypes/culture independence and historical producer equivalence are not newly authenticated.
Counts may be reused across source events. Absence of a full long chain in 100-nt products is not biological absence; no independent replication or new experiment is claimed.
Two exact current-reference TP73 first-junction paths are not returned by this motor-neuron-context junction query; a shared downstream control is returned. The missing output does not resolve which 5-prime route is used and is not evidence of biological non-expression.
Exact reference question
Human TP73 ENSG00000078900, GRCh38 chr1 plus strand. The predeclared first-junction intervals were[3652641,3682332) for ENST00000354437.9 and[3690944,3707548) for ENST00000378288.8; a shared downstream control was[3707791,3722020). Official Ensembl exon records bound these intervals before inference.
Actual output and comparison boundary
One ALL_FOLDS reference PredictInterval/SPLICE_JUNCTIONS request used chr1:[3155806,4204382) and CL:0000100 in-vitro motor-neuron context. In the primary ENCODE channel, both first junctions are unreturned and the shared control has raw signal 0.0120518337935209. All actual channels remain separate in the table. This is output coverage, not a cell-state experiment or a biological absence test.
What changes and the next question
The selected first-junction output cannot resolve the proposed5-prime distinction. A useful next endpoint is source-bound first-exon/TSS coverage or an exact informative5-prime RNA measurement, with its own prespecified readout. Repeating cells or annotations solely to obtain a larger signal would not solve the identity or endpoint issue.
Reference junction predictions; missing values are not zero or biological absence · 9 rows
Predeclared edge
Actual response track
Channel provenance
Raw model value
Output coverage
TP73_201_first
CL:0000100_polyA plus RNA-seq-encode
imputed_exploratory
Not reported
not returned / unscored
TP73_204_first
CL:0000100_polyA plus RNA-seq-encode
imputed_exploratory
Not reported
not returned / unscored
shared_downstream_control
CL:0000100_polyA plus RNA-seq-encode
imputed_exploratory
0.0098545374348759
returned
TP73_201_first
CL:0000100_polyA plus RNA-seq-gtex
imputed_exploratory
Not reported
not returned / unscored
TP73_204_first
CL:0000100_polyA plus RNA-seq-gtex
imputed_exploratory
Not reported
not returned / unscored
shared_downstream_control
CL:0000100_polyA plus RNA-seq-gtex
imputed_exploratory
0.0081053776666522
returned
TP73_201_first
CL:0000100_total RNA-seq-encode
observed
Not reported
not returned / unscored
TP73_204_first
CL:0000100_total RNA-seq-encode
observed
Not reported
not returned / unscored
shared_downstream_control
CL:0000100_total RNA-seq-encode
observed
0.0120518337935209
returned
Interpretation limits
These are current GRCh38 reference predictions, separate from the frozen GRCh37 models and very low-count four-library SMA source context.
ENST00000354437.9 has a 499-aa translation and is not an authenticated TA-alpha/P1 protein. ENST00000378288.8 is annotation-linked to deltaN-alpha; no sample protein or promoter switching is established.
Unreturned edges are unscored, not zero. No isoform abundance, promoter activity, SMA effect or causal variant is inferred.
Catalog and prior response provenance differed for CL:0000100; actual channels from this response are retained. Observed labels describe channel provenance, not a new experimental RNA measurement.
AlphaGenome 2 (AG2): canonical and terminal cryptic STMN2 junction representation
A14 · A26 · Source-linked research context
The reference motor-neuron-context output returns both the canonical STMN2 exon 1-to-2 junction and the published terminal cryptic exon 2a acceptor path. This represents the source-defined sites; it does not predict TDP43-dependent induction or establish cryptic splicing in SMA.
Exact source and coordinate binding
Human STMN2 ENSG00000104435, GRCh38 chr8 plus strand; canonical transcript ENST00000220876.12 / NM_007029.4. Official Ensembl exons bind the canonical intron[79611214,79636801). Melamed et al. report terminal cryptic exon 2a at hg38 chr8:79616822–79617048,1-based inclusive, 227 nt; normalized exon span[79616821,79617048), with cryptic intron[79611214,79616821). The original primary text and annotation were checked before inference.
Actual reference output
One ALL_FOLDS reference PredictInterval/SPLICE_JUNCTIONS request used chr8:[79099719,80148295) and CL:0000100. The primary ENCODE channel returns raw signal 37.4590072631836 for the canonical junction and 2.6184701919555664 for the terminal cryptic edge. All returned channels and their actual provenance remain separate. These raw values are not measurements or a common response scale.
Strongest counterfinding and next question
The primary sources induced cryptic processing by loss of TDP43 regulation; the AG2 reference query contains no such perturbation. The model site is now addressable, but phenotype interpretation requires exact independent source controls. A useful cis followup must use an authenticated published splice-acceptor or regulatory allele with explicit assay context; no allele is invented from a gene name or a larger score.
Reference junction predictions; missing values are not zero or biological absence · 6 rows
Predeclared edge
Actual response track
Channel provenance
Raw model value
Output coverage
canonical_exon1_exon2
CL:0000100_polyA plus RNA-seq-encode
imputed_exploratory
35.547298431396484
returned
terminal_cryptic_exon2a
CL:0000100_polyA plus RNA-seq-encode
imputed_exploratory
3.215216875076294
returned
canonical_exon1_exon2
CL:0000100_polyA plus RNA-seq-gtex
imputed_exploratory
33.08341217041016
returned
terminal_cryptic_exon2a
CL:0000100_polyA plus RNA-seq-gtex
imputed_exploratory
1.910067439079285
returned
canonical_exon1_exon2
CL:0000100_total RNA-seq-encode
observed
37.4590072631836
returned
terminal_cryptic_exon2a
CL:0000100_total RNA-seq-encode
observed
2.6184701919555664
returned
Interpretation limits
The cryptic exon 2a source concerns ALS/TDP43-loss biology. Separate SMA STMN2 overexpression and gene-level RNA findings do not establish that mechanism.
Static reference input cannot perturb TDP43 or reproduce its depletion, proteasome inhibition, transgene delivery or rescue. A nonzero reference signal does not establish measured cryptic RNA in an untreated source cell.
No ratio of canonical/cryptic raw signals is treated as PSI or isoform abundance; no protein, axonal function, treatment benefit or model accuracy is inferred.
Exon2a is terminal through premature polyadenylation. No cryptic-to-downstream splice edge is invented. Exact served weight revision and training overlap remain unbound.
The older variant archive and source-specific measured records remain separately identified in the Evidence Explorer.
AG2 questions across all 44 research axes
All 44 registered axes assessed against retained source-specific work, with targeted RNA/regulatory follow-up. This is not 44 AG2 inferences, exhaustive variant coverage or biological validation.
Source-assessed axes: 44. Axes with current source-linked AG2 model contexts: 9. An axis may need RNA, protein, physiological or source-identity work; a gene name alone is not a model experiment.
No current axis-specific AG2 inference · evidence gap
Existing finding: Retained geometry/contact analyses are computational descriptions only. No biological hit; the full tested insert including its 49-aa prefix and physical material remain unbound; SMA-function relevance is unestablished.
Counterfinding: Existing structure/contact mapping is not wholly negative: ligand-contact and common-core analyses were completed for their fixed inputs, but they do not establish activity, assembly, or SMA durability.
Research question: For a specifically source-bound human SARM1 transcript/variant, does a public-reference AG2 model predict a reproducible splice or expression consequence at the exact locus? This is only useful after an exact locus/allele or a narrowly defined transcript-regulatory question is pinned; it cannot answer NADase activity or axonal degeneration.
Next evidence: Identify an authoritative complete tested-insert/material record; without it no construct-linked model question is executable.
This is a source-based research question, not an experimental result or a claim of model efficacy.
No current axis-specific AG2 inference · evidence gap
Existing finding: AMP/ADP neutral all-atom diagnostics corrected atom-name transfer and preserved heavy coordinates exactly (0.0 Å in-memory change); four references remain computational diagnostics. Separate RNA event context is one GM03813 fibroblast line with shared controls and JC/JCEC not independent contrasts. No flux/function conclusion.
Counterfinding: The source RNA event context provides splice-level information but has one cell line and unresolved sample/producer provenance; it does not measure mitochondrial respiration.
Research question: If a specific OXPHOS gene and source-defined transcript event is selected, does AG2 reference sequence support the observed junction topology in the named cell context? This would not establish respiratory-chain flux or explain the current producer/cell-line provenance gap.
Next evidence: Obtain the archived launch/script receipt for exactly the two AMPK input constructs;
This is a source-based research question, not an experimental result or a claim of model efficacy.
No current axis-specific AG2 inference · other method needed
Existing finding: Human motor-neuron PGK1 RNA rows conflict: GSE69175 log2FC +1.25283, q=0.00091257 versus GSE108094 +0.120896, q=0.954189 plus a separate NOTEST row. PXD015115 lists 77 raw files and one inaccessible combine.xlsx; PXD060094 lists 20/20 iHeps and no iMN. No protein contrast or activity result.
Counterfinding: The second RNA source is near-null and the purported iMN proteome is not present in the reviewed PXD060094 inventory; neither establishes PGK1 protein or glycolytic flux.
Research question: After retrieving the existing PXD015115 combine.xlsx and sample key, does its explicitly identified human motor-neuron PGK1 protein row agree in direction with either RNA source? AG2 cannot resolve protein abundance or enzyme flux.
Next evidence: Retrieve the already identified PXD015115 combine.xlsx plus its sample/reporter/T1 key; if unavailable, record the exact missing file rather than model.
This is a source-based research question, not an experimental result or a claim of model efficacy.
No current axis-specific AG2 inference · other method needed
Existing finding: Pinned native MW150 P11 descriptive arithmetic was independently reproduced and accepted as descriptive only; it does not resolve administered lot, salt/formulation or transfer to SMA.
Counterfinding: The bounded animal-level descriptive contrast exists, but no authenticated administered material or exposure link supports interpreting it as MAPK14 inhibition or broad efficacy.
Research question: No AG2 question can answer the present dosing/material gap. Only after exact administered compound identity is recovered could a separate source-bound MAPK14 transcript/variant question be considered; it would not predict inhibition or efficacy.
Next evidence: Recover the actual administered material/lot/formulation record and animal-cohort link; this is a source/chemistry question, not AG2.
This is a source-based research question, not an experimental result or a claim of model efficacy.
No current axis-specific AG2 inference · other method needed
Existing finding: The deposited P10 whole-liver protein matrix is already checksum-verified and analyzed. FECH/HMBS are higher in the primary SMA subset, while redox markers show mixed directions; SMA-5 remains excluded. The separate accepted heart Nfe2l2 mRNA negative remains unchanged.
Counterfinding: The protein panel is mixed, from whole liver, and is not the separate heart Nfe2l2 mRNA result, a validated pathway response, or independent replication; missing protein rows and sample mapping limit interpretation.
Research question: Which authenticated animal/specimen identities link or distinguish the liver and March cross-organ proteomics sources?
Next evidence: Bind a real specimen crosswalk or exact deposition accession; no matrix reacquisition or repeated analysis. Reagent tables and file suffixes are not animal identity.
This is a source-based research question, not an experimental result or a claim of model efficacy.
Current AG2 context available · reference result with open endpoint question
Existing finding: ML006/SMA was incomplete and is excluded from the comparison. Its partial data do not resolve TP73 isoform support. A current-reference AG2 query returns a shared downstream control but neither predeclared first-junction path.
Counterfinding: Most relevant negative/completion bound is that ML002 has no support in the fixed windows; the incomplete ML006 cannot be counted in either direction, and existing windows are not isoform-specific proof. Unreturned model junctions are unscored, not biological absence or an isoform switch.
Research question: Can an exact first-exon or transcription-start-site coverage endpoint distinguish the two source-bound TP73 paths when junction-list output does not?
Next evidence: Predeclare the first-exon/TSS regions and a suitable output endpoint using the same exact reference identities; retain the difference from frozen low-count source models. No TA-alpha protein or promoter-switching claim.
This is a source-based research question, not an experimental result or a claim of model efficacy.
No current axis-specific AG2 inference · evidence gap
Existing finding: A bounded native source RFD4 feature-processing check passed 13 checks; 384 target atoms were internally conditioned and zero absolutely fixed. No generated binder geometry, model output, or source/resource qualification is established.
Counterfinding: The 384-atom internal conditioning and no absolute restraints do not establish a credible complex or biological interaction; source GSM/library-to-preparation/muscle/animal identity remains open.
Research question: No useful AG2 question until the source sample/preparation map is bound; the existing question is whether an input/library belongs to the claimed muscle model, which is provenance, not sequence prediction.
Next evidence: Recover the exact GSM/library-to-preparation, muscle pool and animal map; no model call is warranted before that source binding.
This is a source-based research question, not an experimental result or a claim of model efficacy.
No current axis-specific AG2 inference · new reference question
Existing finding: The retained Ottesen event evidence concerns MAPK8 in one GM03813 cell line with shared DMSO controls: HiR1000 ΔPSI −0.159, FDR about 0.0200 (JC) / 0.0182 (JCEC); LoR ΔPSI −0.069, FDR 1.0. This is not a dose response or independent replication.
Counterfinding: The nominally relevant event is MAPK8, not MAPK10; one cell line/shared controls and differing JC/JCEC modes constrain any pathway interpretation.
Research question: For the exact source-bound MAPK8 event, does a public-reference AG2 prediction preserve the observed splice junction direction across transcript contexts? It cannot establish MAPK10 involvement, treatment response, kinase activity, or explain JC/JCEC differences.
Next evidence: Use the already retained exact MAPK8 event coordinates/reference and ask one public-reference junction-topology question; no drug-effect claim.
This is a source-based research question, not an experimental result or a claim of model efficacy.
No current axis-specific AG2 inference · other method needed
Existing finding: Three PBA records remain distinct: sodium 4-PBA 10 µM NIH3T3-SMN2/SmnRNAi nuclear-cell normalized value 0.71 vs vehicle; free-acid 4-PBA PubChem AID1458 engineered HEK293 reporter inactive; sodium PBA 2 mM SMA iPSC motor neurons (Type I 1-38G, Type II 1-51N) about 25% higher full-length SMN mRNA but no SMN protein increase and no reliable PBA-specific survival effect. These do not establish HDAC2-specific response; DOI 10.1016/j.stem.2015.08.003.
Counterfinding: The engineered reporter is inactive and the SMA iPSC study did not show SMN protein or reliable survival increase; broad PBA exposure is not proof of HDAC2 target engagement.
Research question: No repeat of the completed source reconciliation. A public-reference HDAC2/SMN2 sequence question would not test PBA response; the discriminating next question requires matched exact reagent and HDAC2 engagement plus protein/function measurement.
Next evidence: Require the exact authenticated reagent identity plus matched HDAC2 engagement/protein/function evidence; AG2 is not a substitute.
This is a source-based research question, not an experimental result or a claim of model efficacy.
No current axis-specific AG2 inference · other method needed
Existing finding: Three distinct source UUIDs map to PMID 31851921/title; one retained date matches PubMed and two anomalous dates remain unexplained. The 9/11/10 linked-claim counts are per-row API counts, not independent papers or experiments.
Counterfinding: The date anomalies and UUID multiplicity show bibliographic records cannot be treated as independent evidence; contact geometry is not channel activity.
Research question: No AG2 question addresses the open issue. The existing required resolution is whether a named VAP/Kv2.1 construct and runner preserve the exact SEP-tagged construct and compatible physical parameterization; this requires construct/model provenance or cellular measurement.
Next evidence: Recover exact deployed runner/parser/version, SEP-containing post-enrichment input, and run-linked output; absent that, retain the source identity gap.
This is a source-based research question, not an experimental result or a claim of model efficacy.
No current axis-specific AG2 inference · other method needed
Existing finding: The Taiwanese-SypHy study directly measured stimulus-dependent mitochondrial calcium responses and pyruvate+lactate fuel replacement with sensor/potential controls. It does not identify selective MCU rescue or astrocytic lactate transfer; SypHy cannot fully separate endocytosis from reacidification. Differently normalized calcium values were not combined as a dose interaction.
Counterfinding: The functional calcium/fuel observations are real but do not isolate MCU-specific rescue or demonstrate astrocyte-to-neuron transfer; SypHy endpoint ambiguity remains.
Research question: No AG2 model can answer the existing functional distinction. The open item is source provenance of the two MCU–EMRE outputs; sequence prediction would not identify their producing seeds or demonstrate calcium flux.
Next evidence: Recover the original manifests for the two specifically named MCU–EMRE outputs; if unavailable, preserve that exact producer-link gap.
This is a source-based research question, not an experimental result or a claim of model efficacy.
No current axis-specific AG2 inference · evidence gap
Existing finding: Chen and Kang source records remain separate. Chen supports truncated fragment/vector labels and deletion design but not complete sequence/accession/version/species/tag/linker or exact plasmid-source identity. Historical labels do not prove full insert. Functional endpoint review is complete; no qualifying new construct-source record was found.
Counterfinding: The fragment/deletion design is documented, but sequence/construct incompleteness prevents assigning a transcript or full-length sequence model to the tested construct; transport endpoint is separate.
Research question: Once a paper-specific full insert/clone sequence is recovered, a public-reference SNPH transcript/context comparison could check sequence identity; it cannot establish transport anchoring or cellular action.
Next evidence: Obtain the paper-specific, versioned full insert/plasmid record for the exact construct; only then assess whether reference sequence comparison is meaningful.
This is a source-based research question, not an experimental result or a claim of model efficacy.
Existing finding: 15 Ottesen source runs span five GM03813 treatment groups but one fibroblast line, not 15 donors; one diagnostic site, separate MAPQ/NH marginals, reads not molecules, and no transcript phasing or treatment-effect inference. MS023 continuation adds three libraries and 88,480 source rows/5,168 event-filter contrasts; Full8 LOO strict sign reversal in 451/1,876 events (median absolute index difference 0.00832 for those). Biological independence is unresolved; no confirmed hit.
Counterfinding: The leave-one-out sign reversals show substantial event-level instability, while single-cell-line source design and unbound biological units limit interpretation; no SMN protein/phenotype link is established.
Research question: Only if an exact existing SMN1/SMN2 variant or source event remains untested in the AG2 graph, ask whether public-reference prediction distinguishes its splice junctions. Current data instead need animal/litter/pool/culture assignments and RNA-to-protein/phenotype links;
Next evidence: Verify exact SMN1/SMN2 context IDs against the already retained AG2 source graph; if represented, mark reused, otherwise first obtain the biological-unit map and RNA-to-protein/phenotype linkage.
This is a source-based research question, not an experimental result or a claim of model efficacy.
Current AG2 context available · reference result with open cis source question
Existing finding: GSE290979 adds 31 genes, normalization sensitivity and eight markers; STMN2/CHAT is lower in six line comparisons while STMN2/RBFOX3 is mixed. There is no cell-composition correction or MN-specific effect. Sixteen conflicting treatment labels remain blocked; GSE290980 is same study with cell annotation unresolved. AAV9 insert species remains unresolved. The new generic motor-neuron reference output returns both canonical and terminal cryptic STMN2 junctions.
Counterfinding: Normalization/marker choice changes the context and mixed RBFOX3 results counter a simple MN-specific signal; treatment labels cannot support a contrast and AAV material origin is unresolved. This static DNA input cannot model TDP43 depletion or establish cryptic splicing in SMA.
Research question: Which exact source-authenticated splice-acceptor or regulatory allele can test the now-addressable cryptic site, with an explicit assay context and interpretable controls?
Next evidence: Bind the exact published cis allele/reference base and source assay before another model request; do not invent a mutation or equate raw signals with PSI.
ALS/TDP43-loss source context remains separate from SMA overexpression; no trans-state induction or rescue is predicted.
Existing finding: MAPK12 p38-gamma reference predictions represent five source-defined junctions. Existing GM03813 RNA outputs and ordered100nt alignment patterns support the nonexclusive exon14-to15-to16 partial chain in all six libraries; no full transcript or treatment response established.
Counterfinding: JC and JCEC use the same libraries and give different qualification; the central SE event has FDR1.0. Generic dermal reference outputs are not treated GM03813. No full four-exon chain, protein or kinase consequence follows.
Research question: Which exact long-range RNA or linked-fragment evidence can distinguish complete MAPK12 transcript structures beyond the observed14-to15-to16 partial chain?
Next evidence: Reuse the accepted five-edge and partial-pattern results; identify a genuinely informative longer molecule/fragment product before a full-chain claim.
This is a source-based research question, not an experimental result or a claim of model efficacy.
Current AG2 context available · other method needed
Existing finding: One stored SMN-C2/RNA Row13 trajectory (1000 frames; 367 base-near, 0 canonical base-H-bond; first 589 frames only 4 base contacts). Row13/Seq9 vs Row12/Seq6 differ at two bases and begin from different states/candidate sets.
Counterfinding: Single trajectories, correlated frames, fixed +1 microstate, constructed RNA and uncertain force-field terms; no causal sequence selectivity or experimental binding inference.
Research question: Can the exact retained Row13/Seq9 versus Row12/Seq6 sequence difference be tested in a matched, prespecified splice-reporter dataset, with matched starting state and construct context?
Next evidence: Missing original construct sequence/boundaries plus matched reporter readout for both constructs in the same cell/model context; AG2 cannot interpret ligand/RNA binding trajectories as splicing efficacy.
This is a source-based research question, not an experimental result or a claim of model efficacy.
Existing finding: Accepted structural stoichiometry retained 820 source-bound thresholds, 518 chain-specific pairs, 302 under 4 Å; structures are T2R α/β complexes with rat STMN4/RB3-SLD, not STMN1. Separate Oct 25 SYT2/NCALD RNA/literature context.
Counterfinding: Structural contact counts do not demonstrate human STMN1 binding or SMA contracture mechanism; no complete STMN1 constructs/native inputs.
Research question: Can an exact human STMN1 construct and native oligomer/input set be bound before any structural comparison is interpreted?
Next evidence: Missing complete human STMN1 construct sequence/isoform and matched native input identity; do not duplicate PLS3 variant modeling assigned elsewhere.
This is a source-based research question, not an experimental result or a claim of model efficacy.
No current axis-specific AG2 inference · evidence gap
Existing finding: MIB1 WT vs C1009S HEK293T [35S] pulse-chase: about 4 h vs 2 h, 3 experiments. CHEMBL4068807 maps to Rietz 4c (0.28±0.03 µM; 190±32% SMN-luciferase), not direct protein/artifact-control evidence. HSPA8 peptides/chimeras separately identified.
Counterfinding: Open source-unit and processing gaps remain for Harmine/PIP6A RNA/protein datasets; reporter is indirect and no exact mechanism/isoform prediction follows.
Research question: Which GSMs entered the published 11-sample GSE150510 DESeq2 analysis, and what original reference/count recipe generated the mixed gene/transcript-form rows?
Next evidence: Missing author inclusion list keyed to GSMs plus original RNA reference/quantifier/aggregation recipe; no AG2 rerun can resolve sample inclusion or count construction.
This is a source-based research question, not an experimental result or a claim of model efficacy.
No current axis-specific AG2 inference · evidence gap
Existing finding: INPP5F loss improves one injury-regeneration context but impairs selected N2A recycling; WT rescue and C458S differ. GSE145146 has n=1 DMSO vs n=3 treated; INPP5F CPM 22.68 vs 22.39/23.69/23.37.
Counterfinding: Small unbalanced RNA libraries cannot support drug-response or mechanistic differential-expression claims; exact historical Hsu construct/insert remains unresolved.
Research question: What exact historical Hsu plasmid sequence and expression insert/junctions were used, and do they encode the same INPP5F construct as the current reference?
Next evidence: Missing accession-bound historical plasmid or full construct sequence and readable primary Figure 1/source; avoid guessing mutant identity or treating AG2 as construct-sequence validation.
This is a source-based research question, not an experimental result or a claim of model efficacy.
No current axis-specific AG2 inference · other method needed
Existing finding: Kv2.1 model covers 21,870 simulated cells; complete-loss firing rate is 0.308–0.500 of WT while AP half-width is 0.98–1.20. Separate SCS fit uses 3 SMA participants and 7 external controls.
Counterfinding: Model outputs and tiny heterogeneous SCS sample do not establish channel-specific measured human SMA physiology; exact KCNC4 pore structure/construct missing.
Research question: Can an authoritative human KCNC4/Q03721 pore-containing structure or construct be bound without substituting KCNC1 7PQT?
Next evidence: Missing authoritative exact human KCNC4 pore-containing structure or residue/construct map and assembly; this is a structure identity question, not RNA regulation.
This is a source-based research question, not an experimental result or a claim of model efficacy.
Current AG2 context available · other method needed
Existing finding: GSE2566: n=3 paired rat delayed-peroneal reinnervation samples with opposing Perp probes. GSE252128: 8 SMA-labelled human surgical muscle vs 7 scoliosis controls, PERP TPM means 7.8146 vs 2.1297; all 15 one-sample omissions positive but ranges overlap. P7 mouse array/RNA-seq are same study.
Counterfinding: Different species/tissues and partly repeated cohorts; overlapping values and probe disagreement limit transfer; no direct NMJ localization/binding demonstrated.
Research question: Can source-level sample/animal identity and exact PERP probe/feature mapping distinguish independent evidence from reused cohorts and opposing probe behavior?
Next evidence: Missing exact original sample crosswalk and probe/feature annotation/source for PERP; AG2 is not a substitute for cohort or localization evidence.
This is a source-based research question, not an experimental result or a claim of model efficacy.
No current axis-specific AG2 inference · other method needed
Existing finding: Three MuSK-Ig3 BindCraft designs (65/67/91 aa) had 71/102/111 interchain pairs <4 Å in rigid 9NWG projection. Separate RFD4 70-aa B-chain has one Cys and no N-X-S/T sequon; nearest target Cys 16.7589 Å. Boltz2 template-conditioned run had no <4 Å interchain pairs.
Counterfinding: Model/design outputs conflict and rigid projection is not binding validation; Crk/CrkL paralog and construct context remain distinct.
Research question: Do accepted explicit-pY contact poses overlap exact experimental Crk/CrkL-SH2 reference contacts for the same phosphosite/construct?
Next evidence: Finite source comparison of retained explicit-pY poses against exact experimental Crk/CrkL-SH2 reference contact source; no AG2 call.
This is a source-based research question, not an experimental result or a claim of model efficacy.
No current axis-specific AG2 inference · other method needed
Existing finding: Eight raw Agrin–LRP4 designs plus eight inverse-folded sequence records represent eight designs/16 sequence records, all 80 aa. All backbones overlap eligible AGRN atoms (minimum 0.102–0.469 Å); no pose nominated. Separate context-model attempt SIGBUSed before output.
Counterfinding: Severe overlap prevents nomination; failed model has no result, and structural geometry does not establish binding/stability.
Research question: Can the exact source atom/model/chain selection and overlap mask be independently bound for all eight designs before any redesigned pose is considered?
Next evidence: Use retained structures/source map for finite source atom-identity and overlap-mask reconciliation; no AG2 call,
This is a source-based research question, not an experimental result or a claim of model efficacy.
No current axis-specific AG2 inference · evidence gap
Existing finding: SAPPHIRE FDA table: age 2–12, apitegromab 10 mg/kg vs placebo n=53/50, adjusted HFMSE difference 2.2 (95% CI 0.49–3.95), nominal p=.0121. Separate from primary 20 mg/kg and pooled-dose contrast.
Counterfinding: Prespecified treatment×SMN-background/age interaction outputs remain absent; one dose subgroup cannot establish mechanism or general efficacy.
Research question: What are the original prespecified treatment-by-SMN-background and age interaction estimates for the retained SAPPHIRE endpoints?
Next evidence: Missing original prespecified interaction output/table with model specification; AG2 cannot answer clinical subgroup interaction.
This is a source-based research question, not an experimental result or a claim of model efficacy.
No current axis-specific AG2 inference · other method needed
Existing finding: AF3 series 30/30 identity/hash checked; U1-C/RNA ± SMN-C5 PAE medians differ but contact directions vary and severe non-Zn overlaps 7/15 vs 1/15. Source14 fixes second combo library SRR19737679/GSM6253090/SRX15783611/SAMN29180978; lower-skip pattern does not persist; no consistent add effect/interaction.
Counterfinding: Model contact/PAE is not affinity or splice efficacy; combination sample does not reproduce a consistent add effect; risdiplam vs SMN-C5 affinity/activity mapping unresolved.
Research question: Can exact original matched RNA-binding affinity and cellular splicing potency measurements be recovered for risdiplam and SMN-C5 under the same RNA/assay conditions?
Next evidence: Missing exact primary assay pair with matched construct, condition and numeric affinity/activity endpoint; AG2 is unsuitable for compound affinity or RNA–protein–ligand geometry.
This is a source-based research question, not an experimental result or a claim of model efficacy.
Current AG2 context available · reference result with open cis source question
Existing finding: Pagliari S4D raster means show similar GFAP/IBA1 for SMA+null and SMA+Stmn2, n=3/group. GSE59506 counters uniform axonal mRNA decrease: both Stmn2 probes higher in Smn-shRNA vs WT/empty in soma and axon; one axonal contrast +0.249 and +0.709. The new generic motor-neuron reference output returns both canonical and terminal cryptic STMN2 junctions.
Counterfinding: Pagliari lacks stable animal IDs; raster means small n; probe-level direction is not axonal maintenance function. This static DNA input cannot model TDP43 depletion or establish cryptic splicing in SMA.
Research question: Which exact source-authenticated splice-acceptor or regulatory allele can test the now-addressable cryptic site, with an explicit assay context and interpretable controls?
Next evidence: Bind the exact published cis allele/reference base and source assay before another model request; do not invent a mutation or equate raw signals with PSI.
ALS/TDP43-loss source context remains separate from SMA overexpression; no trans-state induction or rescue is predicted.
No current axis-specific AG2 inference · evidence gap
Existing finding: Adami P11 Δ7 SVZ NSCs: nominal 0.5 μM “Cur” changes NRF2 nuclear/cytoplasmic ratio 1.522±.023 to 1.598±.021 (n=4, p=.0464); NQO1 nonsignificant, SMN remains below WT. MTT text says 16% while means imply 21.9%; Ottesen DNM1L context is shared A02/A27/A33, not NRF2 evidence.
Counterfinding: Formulation/lot uncertain; effect is nominal and NRF2 reporter lacks downstream NQO1 significance; DNM1L must not be used as NRF2 evidence.
Research question: Can the original Adami supplement and methods resolve compound identity/formulation and reconcile the MTT percent with the reported means?
Next evidence: Missing readable original supplement/compound specification and raw MTT values or calculation definition; no AG2 variant prediction warranted.
This is a source-based research question, not an experimental result or a claim of model efficacy.
Counterfinding: No author QC/cluster/lineage metadata in retained GEO raw archive; proxy strata are not author cell identities; zebrafish/mouse/human contexts differ.
Research question: Can GEO or authors supply GSM+barcode-bound QC/cluster labels and explicit pMN→newborn MN birth/lineage labels for the same cells?
Next evidence: Missing barcode-linked author QC/cell-state and birth/lineage metadata; existing threshold sensitivity and marker codetection are complete and do not satisfy original lineage contrast.
This is a source-based research question, not an experimental result or a claim of model efficacy.
No current axis-specific AG2 inference · evidence gap
Existing finding: Wu 2025 P4-heart miR-34a-5p qPCR (~4-fold) and stronger FISH; six-run prefix-CPM is directionally similar but not sample-linked. miR-181a-5p lacks quantitative validation.
Counterfinding: qPCR n=10 is not mapped to the six sequencing runs; organ, cohort, and intervention linkage remains open.
Research question: Can any retained source manifest bind each P4 heart qPCR sample to the six RNA runs and exact intervention/genotype?
Next evidence: Missing input: per-sample crosswalk linking qPCR n=10, six RNA runs, heart tissue, genotype and intervention; without it no AG2 contrast is interpretable.
This is a source-based research question, not an experimental result or a claim of model efficacy.
No current axis-specific AG2 inference · other method needed
Existing finding: Fifteen selected arms across three papers show dose- and assay-specific outcomes; no formal interaction or general synergy is established.
Counterfinding: No matched PLS3-only arm in one mouse comparison; Riessland untreated survival labels conflict; some outcomes are within common-ASO arms.
Research question: Do original datasets contain complete matched factorial SMN-treatment × modifier-treatment arms with comparable independent units and endpoints?
Next evidence: Missing input: source-bound complete matched arms, animal/sample IDs, dose and endpoint definitions sufficient for a treatment-interaction analysis.
This is a source-based research question, not an experimental result or a claim of model efficacy.
No current axis-specific AG2 inference · other method needed
Existing finding: SAPPHIRE treatment-by-SMN-background/age interaction remains unquantified; a corrected human GM03813C MSTN event query returned zero matching rows in its fixed prefiltered set.
Counterfinding: Zero of 9,778 prefiltered rows is not a global absence; the earlier mouse gene ID was wrong and no interaction estimate exists.
Research question: Can exact tested myostatin material and original intervention arms be bound to a muscle endpoint that permits a pre-specified interaction contrast?
Next evidence: Missing input: exact myostatin intervention identity plus complete matched SAPPHIRE arm-level outcomes and SMN-background/age labels.
This is a source-based research question, not an experimental result or a claim of model efficacy.
No current axis-specific AG2 inference · other method needed
Existing finding: Kim 2017 reports selected proximal NMJ/muscle/survival benefits in low-SMN mice with Hb9 chicken Z+Agrin, while motor-neuron/vGlut1 and distal innervation were unchanged and whole-cord SMN was not raised.
Counterfinding: DUET P14 sex-stratified intervals disagree by method; DUET also adds Cas9 and doubles vector genomes, so arms are not isolated single-factor contrasts.
Research question: Are source-complete matched intervention arms available to isolate the construct and developmental-stage contribution to NMJ endpoints?
Next evidence: Missing input: complete source-linked group assignments and independent-animal endpoint table for comparable construct arms.
This is a source-based research question, not an experimental result or a claim of model efficacy.
No current axis-specific AG2 inference · evidence gap
Existing finding: Grandi 2024 muscle data show RNA/protein/mtDNA differences without respirometry or before-after treatment; James2024 redox contrasts change direction by cell-line origin; proteome sensitivity found no corrected hits.
Counterfinding: PBMC mtDNA is a separate cohort; SMN protein did not differ in the measured comparison; MitoSOX subgroup directions oppose and do not prove batch effects.
Research question: Can retained sample and assay metadata bind the muscle, PBMC, and cell-line measurements to their actual cohorts, normalization, and intervention status?
Next evidence: Missing input: exact sample-to-cohort/assay/normalization crosswalk and any matched respirometry or intervention values; AG2 transcript predictions would not resolve these endpoints.
This is a source-based research question, not an experimental result or a claim of model efficacy.
No current axis-specific AG2 inference · evidence gap
Existing finding: NCALD structural projection is a computational geometry check, not a measured calcium-bound complex; other PERP/FAS4, P4 mouse and CDK4 drug-window evidence remains context-specific and nonvalidating.
Counterfinding: Clathrin sources disagree on calcium direction; only 69 somatomotor nuclei among six human donors; P4 cells are pooled littermates; CDK4 windows are modeled and lack validated complexes/cell assays.
Research question: For any proposed NCALD RNA question, is there a source-bound exact allele or transcript interval and a matched human motor-neuron measurement beyond the existing structural and context analyses?
Next evidence: Missing input: a specific source-bound NCALD allele/transcript question with exact motor-neuron RNA endpoint; absent that, no new AG2 request is justified.
This is a source-based research question, not an experimental result or a claim of model efficacy.
No current axis-specific AG2 inference · other method needed
Existing finding: Corrected Fig. 7D endpoint is proprioceptive VGluT1+ inputs on ChAT+ L2 motor-neuron somata at P11, not motor-neuron counts; the delayed SMN-C3/MW150 combination did not significantly increase that endpoint.
Counterfinding: Direction-only +1.297619 normalized units and 25/25 leave-one-out results are descriptive; survival RMST concerns a different comparison; Munc13 data lack independent replicate pairing.
Research question: Can source records link the Munc13 single-cell values to independent biological replicates and paired controls for a valid contrast?
Next evidence: Missing input: original Munc13 sample-to-replicate/control mapping; AG2 sequence scoring cannot substitute for NMJ endpoint evidence.
This is a source-based research question, not an experimental result or a claim of model efficacy.
No current axis-specific AG2 inference · evidence gap
Existing finding: SLC6A6 muscle RNA contrasts are source-specific; post-hoc human/mouse values and treatment contrast do not supply paired taurine measurements. PERP-CSAD single-library relation is descriptive; LRRC58/CDO1 synthesis differs from SLC6A6 transport.
Counterfinding: OMIX006777 is not released; the required same-person muscle RNA/taurine and side/aliquot mapping is unavailable; PXD060060 did not resolve treatment reference arm.
Research question: Does a retained accession or source table bind SLC6A6 muscle RNA and taurine concentration in the same person and aliquot, with treatment/genotype recorded?
Next evidence: Missing input: released source data or exact same-person sample/side/aliquot crosswalk with taurine values; do not run an AG2 reference prediction as a proxy for this paired test.
This is a source-based research question, not an experimental result or a claim of model efficacy.
No current axis-specific AG2 inference · other method needed
Existing finding: Existing P7 mouse triceps matrix adds Dnajc5 values with nonsignificant contrasts; treatment interaction is near zero, and HSPA8 control studies have unresolved construct identity.
Counterfinding: The RNA result is small and statistically unresolved, not an independent cohort; DnaJ chimera/preprint identities and HSPA8/NEF construct/run pairing remain unbound.
Research question: Can the exact published and tested HSPA8/NEF constructs, chimera junctions, tags, and raw run/well pairing be recovered to resolve the control identity before any calibrated comparison?
Next evidence: Missing input: exact chimera junctions and construct/tag sequences plus source-bound raw run/well pairing; no new AG2 run needed.
This is a source-based research question, not an experimental result or a claim of model efficacy.
No current axis-specific AG2 inference · other method needed
Existing finding: GSE320428 arrays and sample labels are bound; PERP/p53 association direction is robust to fixed omissions, while PERP/microglia association changes sign across omissions.
Counterfinding: Biological replicate-to-animal/pool mapping is missing; no new p/q values or CIs; the estimand and error model are not specified.
Research question: What estimand and covariance/error model should govern the existing PERP/p53 and PERP/microglia contrasts, given unresolved replicate grouping?
Next evidence: Next step: specify the estimand and error model from the documented biological sample structure, using the pinned source review; only then decide whether existing data support additional inference.
This is a source-based research question, not an experimental result or a claim of model efficacy.
No current axis-specific AG2 inference · other method needed
Existing finding: D10/P8 source Oligo overlaps one possible 20-nt window in reverse-complement orientation once in SMN2; none in SMN1. Separately, none of 19 saved local guide strings matches four Oligo windows.
Counterfinding: The possible spacer-window hit is only sequence identity; authentic spacer, guide/PAM, construct and editing function remain unverified. Two searches used distinct sequence sets and are not contradictory.
Research question: Can the authentic D10/P8 spacer and complete guide/PAM/construct sequence be source-bound, then compared to the single SMN2 genomic match?
Next evidence: Missing input: original full spacer/PAM and construct sequence with authenticated orientation; AG2 is not an editing-efficiency assay.
This is a source-based research question, not an experimental result or a claim of model efficacy.
No current axis-specific AG2 inference · other method needed
Existing finding: Mapping audit preserves 112,180 RNA/splice rows and 61,166 ambiguous B candidates; GSE290980 has 8 libraries/39,266 barcodes, while the GSE108094 panel has no PERP event and one TP73 SMA value.
Counterfinding: Possible coordinate matches are not confirmed gene identities; two libraries per cell line are not independent donors; historical regulon code mishandled zero/NaN/Inf and the offline fix is not deployed.
Research question: Which documented method version and sign convention produced the historical regulon result, and which source rows were affected?
Next evidence: Missing input: immutable producer/helper source and execution logs for the regulated output; do not use AG2 to validate ambiguous row-to-gene mapping.
This is a source-based research question, not an experimental result or a claim of model efficacy.
No current axis-specific AG2 inference · other method needed
Existing finding: No qualifying primary source establishes direct WT p62-PB1/SMN binding or SMA muscle p-4E-BP1 reduction; 90-nM structure contains mutant SQSTM1 PB1, not WT or SMN.
Counterfinding: Tagged-cell co-recovery and endogenous mouse-motor-neuron co-IP are distinct from purified direct binding; exact gift-plasmid sequence/tags are unavailable; proteome sensitivity had no corrected hits.
Research question: Can exact WT p62/PB1/SMN constructs and a direct binding assay be source-bound, or a qualifying primary p-4E-BP1 SMA skeletal-muscle measurement be identified?
Next evidence: Missing input: complete plasmid/construct/tag sequences and qualified direct assay, or a source-bound relevant SMA muscle phospho-measurement; AG2 cannot answer protein binding.
This is a source-based research question, not an experimental result or a claim of model efficacy.
No current axis-specific AG2 inference · other method needed
Existing finding: The 400-molecule panel matches a dated Enamine kinase set, but selection-pool definition/order remain unknown. GSK3B biochemical, HEK293 reporter, fibroblast and motor-neuron reports are not interchangeable.
Counterfinding: Makhortova fibroblast ligand labels and PDGF result conflict with separate mouse ESC-derived MN report; tested isoform/preparation/dose are not known to match; Ohuchi uses one patient and three clones, not three people.
Research question: Can the original 48,915-member selection pool and selection script/order be recovered, and can exact tested PDGF preparation/dose be tied across the source assays?
Next evidence: Missing input: original pool-selection provenance plus exact ligand/isoform/dose and assay arm crosswalk; do not treat a GSK3B score as SMN stabilization.
This is a source-based research question, not an experimental result or a claim of model efficacy.
Existing finding: Prior source-bound AG2 SMN2 gene-first, variant/splice and regulatory-source contexts are already integrated in the 11-context Explorer; six donor ATAC comparison has no selected peaks and no SMA donors.
Counterfinding: Historical minigene T−255C/C−255T is not physically bound to rs1317747440/NM_017411.4:c.835−260C>T; no supplement link in checked metadata does not prove none exists; transgene context limits locus-specific interpretation.
Research question: Can an original supplement, accession, or complete minigene sequence bind the historical variant to an exact genomic allele/transcript, without repeating existing AG2 contexts?
Next evidence: Missing input: authentic complete minigene sequence or source supplement/accession establishing physical variant identity; no new AG2 call until this identity is resolved.
This is a source-based research question, not an experimental result or a claim of model efficacy.
Existing finding: Prior source-bound AG2 contexts for PLS3 allele/splicing/reference questions are already integrated; NCALD source libraries are same-donor and not genotype-linked, and GSE290441 lacks its FLAIR annotation model.
Counterfinding: BL35 terminal signature appears only in the defining read; discordant-sibling source has no accessible RNA accession or modifier genotype mapping; PLS3 c.367+182C>T remains separate and identity-unresolved.
Research question: Can the original GSE290441 FLAIR GTF/BED/model or an independent genotype-bound human motor-neuron RNA source be obtained for the existing NCALD/PLS3 question?
Next evidence: Missing input: GSE290441 FLAIR GTF/BED/model or independent source accession binding genotype, transcript, and motor-neuron RNA; no duplicate AG2 PLS3 call.
This is a source-based research question, not an experimental result or a claim of model efficacy.