A22 · COMPUTATIONAL STRUCTURE ANALYSIS · 13 SEPTEMBER 2026
Explicit phosphotyrosine changes modeled DOK7–CRKL contact geometry
A controlled 60-model AlphaFold 3 panel compares unmodified DOK7 with pY395, pY405 and double-phosphorylated inputs against the same CRKL construct. The result is a descriptive sensitivity analysis of modeled contact geometry.
Exact molecular panel
Chain A is human DOK7 Q18PE1 residues 380–430. Local positions 16 and 26 correspond to canonical Y395 and Y405. Chain B is human CRKL P46109 residues 1–120. The four arms are unmodified DOK7, pY395, pY405 and pY395+pY405. Each arm contains seeds 1, 2 and 3 with samples 0–4: 15 models per arm and 60 models in total.
Contact is defined as a minimum heavy-atom distance of 4.0 Å or less. Counts are distinct CRKL construct positions. The common-scaffold mask uses the same 12 TYR/PTR heavy atoms. The phosphate-only mask uses P, O1P, O2P and O3P.
Eight site-by-arm results
Cells report minimum/median/maximum partner-position counts across 15 models, followed by models with at least one contact in parentheses.
| Arm | DOK7 site | All site atoms | Common scaffold | Phosphate only |
|---|---|---|---|---|
| Unmodified | Y395 (local 16; TYR) | 0/5/7 (14/15) | 0/5/7 (14/15) | n/a |
| Unmodified | Y405 (local 26; TYR) | 0/3/5 (14/15) | 0/3/5 (14/15) | n/a |
| pY395 | Y395 (local 16; PTR) | 8/8/8 (15/15) | 4/6/7 (15/15) | 4/5/5 (15/15) |
| pY395 | Y405 (local 26; TYR) | 0/1/6 (12/15) | 0/1/6 (12/15) | n/a |
| pY405 | Y395 (local 16; TYR) | 0/0/1 (1/15) | 0/0/1 (1/15) | n/a |
| pY405 | Y405 (local 26; PTR) | 8/8/9 (15/15) | 5/5/8 (15/15) | 4/5/7 (15/15) |
| pY395 + pY405 | Y395 (local 16; PTR) | 2/4/7 (15/15) | 1/3/6 (15/15) | 2/3/6 (15/15) |
| pY395 + pY405 | Y405 (local 26; PTR) | 6/8/10 (15/15) | 4/5/6 (15/15) | 2/5/7 (15/15) |
What changed in the models
The singly phosphorylated site contacted CRKL in all 15 models for pY395 and all 15 for pY405. Contacts admitted only by the additional PTR atoms occurred at pY395 in 15/15 models and at pY405 in 15/15 models. In the double-pY arm, the corresponding values were 13/15 at Y395 and 15/15 at Y405.
CRKL construct position 21 was the most recurrent phosphate-enabled position in the pY395 arm (15/15); positions 21 and 42 were most recurrent at pY405 (14/15 and 12/15). These counts describe only the frozen contact definition and model ensemble.
Interpretation limits
The arms are separate model runs. Shared seed and sample labels do not establish paired random draws or biological replicates. Explicit PTR changes both chemistry and tokenization.
These models do not demonstrate binding affinity, CRKL recruitment, DOK7 activation, neuromuscular-junction function or an SMA treatment effect. The earlier phenylalanine, glutamate and aspartate substitution panel has different chemistry and denominators and is not pooled here.