Use case 4: Assessing the pathological effects of missense mutations
Problem
We would like to use AlphaFold2 to decide on the possible pathological outcome of missense variations. We have a list of mutations with outcome from UniProt humsavar. We download AlphaFold and predict wild type and variant structures on our own computer, without using templates. Then we use other tools to assess stability of the predicted structures.
Initial dataset
We have a list of UniProt ACs from humsavar together with information about variants.
?
We run the original version of AlphaFold without templates on UniProt sequences. Should we restrict our training set?
!
Use full length protein set with the first cut off date (01.05.2018.)
Proposed protocol with BETA
- The original AlphaFold was trained on structures deposited to PDB before 30.04.2018. We don't use templates.
- Go to the BETA download page and select benchmark version. For the benchmark set we use full length proteins. For date we select the closest date after the cutoff, which is 01.05.2018.
- The intersection of the original benchmark set (steps 1) and the filtered set from BETA (step 2) should be used to calculate stability.