Week 5: Protein Design Part II

Context
This week extends protein design with a peptide-binder workflow. You’ll generate short peptides with PepMLM, then model peptide–protein complexes with AlphaFold-Multimer and interpret ipTM scores. :contentReference[oaicite:0]{index=0}

Goals
- Use PepMLM to propose peptide binders for a chosen target (here: SOD1). :contentReference[oaicite:1]{index=1}
- Predict peptide–target complexes with AlphaFold-Multimer and compare ipTM. :contentReference[oaicite:2]{index=2}
- Summarize results and pick candidates for follow-up.
Part A — PepMLM → AlphaFold-Multimer (protein–peptide)
Warning
Read this whole assignment before starting. You’ll need a Hugging Face token and a GPU runtime for Colab. :contentReference[oaicite:3]{index=3}
Step 1 — Set up PepMLM (Hugging Face)
- Model card: PepMLM-650M (ChatterjeeLab) — background, links, and Colab.
https://huggingface.co/ChatterjeeLab/PepMLM-650M :contentReference[oaicite:4]{index=4} - Create an Access Token (Hugging Face): https://huggingface.co/settings/tokens (paste it when the notebook asks).
Visual guide from my notes:




Step 2 — Prepare the target sequence (SOD1)
- Grab human SOD1 sequence from UniProt (P00441) and edit A4V (A→V at position 4).
https://www.uniprot.org/uniprotkb/P00441/entry :contentReference[oaicite:5]{index=5}
Step 3 — Generate candidate peptides
- Run the PepMLM Colab and generate 4 peptides of length 12 for the mutated SOD1.
- Add this known SOD1-binding peptide to your set: FLYRWLPSRRGG (from the literature). :contentReference[oaicite:6]{index=6}
Step 4 — Structure modeling with AlphaFold-Multimer
- Use ColabFold AlphaFold2 notebook (supports Multimer).
https://colab.research.google.com/github/sokrypton/ColabFold/blob/main/AlphaFold2.ipynb :contentReference[oaicite:7]{index=7} - Set
model_type = alphafold2_multimer_v3. Benchmarking supports AF-Multimer for peptide–protein docking. :contentReference[oaicite:8]{index=8} - For each peptide, submit the complex
SOD1Sequence:PeptideSequenceand record ipTM.
Step 5 — Compare and interpret
- Plot ipTM for all candidates (higher ≈ more confident interface) and note interface details (contacts, pose variety).
- Short write-up: 1 paragraph on which peptide(s) you’d pursue and why (ipTM, pose, chemistry). :contentReference[oaicite:9]{index=9}
Tip
Minimal report for Part A
- Table of peptides (PepMLM 3–4 + literature peptide), the edited SOD1 sequence (A4V), and ipTM per complex.
- 2–3 annotated screenshots from your top model(s).
- One-paragraph conclusion + next steps (e.g., mutate peptide residues; re-score with more seeds).
Part B — Final Project (L-Protein mutants)
This is a compute-heavier follow-up for your final project—start early and keep logs of settings, seeds, and runtimes. (Details per course brief.) :contentReference[oaicite:10]{index=10}
Submission checklist
- SOD1 A4V sequence (source + edit noted). :contentReference[oaicite:11]{index=11}
- List of peptides (PepMLM outputs + FLYRWLPSRRGG reference). :contentReference[oaicite:12]{index=12}
- ipTM plot across candidates; top models’ screenshots. :contentReference[oaicite:13]{index=13}
- 1-paragraph summary of findings & next steps.
References / Tools
- PepMLM model card (links to Colab + paper): https://huggingface.co/ChatterjeeLab/PepMLM-650M :contentReference[oaicite:14]{index=14}
- UniProt SOD1 (P00441): https://www.uniprot.org/uniprotkb/P00441/entry :contentReference[oaicite:15]{index=15}
- ColabFold AlphaFold2 notebook (Multimer): https://colab.research.google.com/github/sokrypton/ColabFold/blob/main/AlphaFold2.ipynb :contentReference[oaicite:16]{index=16}
- AF-Multimer peptide docking benchmark (Frontiers in Bioinformatics, 2022): https://www.frontiersin.org/articles/10.3389/fbinf.2022.959160/full :contentReference[oaicite:17]{index=17}