Skip to content

Milestones

List view

  • Protein interaction screening

    Using efficient methods to screen for interaction partners - One vs all - Pairwise - Specify ratio (more advanced, may instead require users setting up the appropriate multimer files to be handles by samplesheet-utils This also includes things like reusing MSAs rather than recalculating for each interaction. This is easier to do in Boltz-style input where each monomer sequence header must be provided with an appropriate MSA file path. AlphaFold2 requires are prior MSA job and then copying those results into a folder structure that's restructured exactly to what AlphaFold2's ` --use_precomputed_msas=true` flag requires. AlphaPulldown uses the .pkl files but they get unwieldy in light of default user individual storage limits

    Overdue by 1 year(s)
    •
    Due by March 28, 2025
    •0/1 issues closed
    0% complete1 open0 closed

  • Easy 'self-service' of proteinfold web interface

    This includes: - "Upload FASTA files and hit go" (Samplesheet Nathan is working on - Concise and insightful user documentation on appropriate program choice - Publishing computed structures as they finish, not being stuck in user-hidden .proteinfold directories - Adapted On Demand job title cards than link to where the results folder in on the cluster filesystem - A 'summary report' of relevant metrics (mean pLDDT, iPTM, ...) from MultiQC processing of output files

    Overdue by 1 year(s)
    •
    Due by April 4, 2025
    •3/4 issues closed
    75% complete1 open3 closed

  • Highly efficient AlphaFold3 reimplementation (Boltz + MMSeq-GPU)

    Provide the users a default, super efficient way to run AlphaFold3 generation structure predictions. Boltz + MMSeqs-GPU avoids all the intellectual property complications of AlphaFold3 licence, and should be on order of 100x more efficient on current hardware.

    Overdue by 1 year(s)
    •
    Due by February 28, 2025
    •1/2 issues closed
    50% complete1 open1 closed