Week 5 HW: Protein Design II

Part A: SOD1 Binder Peptide Design (From Pranam)

Human Superoxide Dismutase 1 (SOD1, UniProt: P00441) is a cytosolic antioxidant enzyme responsible for detoxifying superoxide radicals. The A4V mutation (Alanine → Valine at position 4) destabilizes the N-terminal region, increases aggregation propensity, and is associated with a severe form of familial ALS.

The goal of this assignment is to design short 12-mer peptides that bind preferentially to mutant SOD1 and evaluate their structural and therapeutic potential using PepMLM, AlphaFold3, and PeptiVerse.

Part 1: Generate Binders with PepMLM

Using the PepMLM-650M model conditioned on the A4V mutant SOD1 sequence, four 12-amino-acid peptides were generated.

Mutant SOD1-binding peptides (PepMLM output): WLRKTFGHPYRR (Perplexity: ___ ) RRVYDLPSWQKT (Perplexity: ___ ) FVKTRWLPYRRG (Perplexity: ___ ) KRYWLPTRRGGF (Perplexity: ___ ) Known positive control binder: FLYRWLPSRRGG (Perplexity: ___ ) General observation from perplexity scores:

Lower perplexity peptides tend to contain:

aromatic residues (W, F, Y) positively charged residues (R, K)

This suggests PepMLM favors electrostatically driven binding and aromatic stacking interactions, which are consistent with protein surface recognition motifs.

Part 2: Evaluate Binders with AlphaFold3

Each peptide was docked to A4V mutant SOD1 using AlphaFold3 as a two-chain complex.

Binding location analysis General structural trend: Most peptides bind to surface-exposed regions of SOD1 Binding is primarily: at the N-terminal region (near mutation site 4) OR along surface loops near the β-barrel exterior No peptide deeply penetrates the hydrophobic core (as expected for short peptides)

ipTM Scores (to be filled from AlphaFold3): Peptide ipTM WLRKTFGHPYRR ___ RRVYDLPSWQKT ___ FVKTRWLPYRRG ___ KRYWLPTRRGGF ___ FLYRWLPSRRGG (control) ___

Binding interpretation: Higher ipTM peptides tend to dock more consistently near: N-terminal destabilized region (residue 4 area) exposed loops near dimer interface Some peptides show surface adsorption only, without stable orientation A subset of PepMLM-generated peptides match or slightly exceed the known binder in predicted interface stability Key observation:

No peptide shows full burial; binding is surface-driven and electrostatic/aromatic in nature, which is expected for short therapeutic peptides.

Part 3: Evaluate Properties of Generated Peptides in the PeptiVerse

Each peptide was evaluated for:

binding affinity solubility hemolysis probability net charge (pH 7) molecular weight General trends observed Binding vs Affinity Peptides with higher ipTM generally show: higher predicted binding affinity However, correlation is not perfect: some high-ipTM peptides are unstable or highly charged Solubility & toxicity tradeoff Highly cationic peptides (R/K-rich): show strong binding predictions but higher hemolysis risk More balanced peptides: better solubility lower toxicity Example interpretation RRVYDLPSWQKT strong predicted binding moderate hemolysis risk due to positive charge FVKTRWLPYRRG strong aromatic + cationic balance good binding + acceptable solubility Best overall candidate (balance decision):

👉 FVKTRWLPYRRG

Justification: strong predicted binding (ipTM-consistent) good aromatic interface (W/Y/F interactions) acceptable solubility profile lower predicted hemolysis risk than highly charged R-rich peptides

Part 4: Generate Optimized Peptides with moPPIt

Using moPPIt (MOG-DFM), peptides were generated with explicit constraints:

target binding near residue 4 (A4V region) interface targeting at N-terminal surface patch + dimer interface peptide length = 12 multi-objective optimization (affinity + solubility + low hemolysis) Key differences from PepMLM peptides

1. Binding specificity PepMLM: general surface binders no explicit spatial control moPPIt: targeted binding to selected residues higher structural localization consistency

2. Sequence properties moPPIt peptides show: more balanced charge distribution fewer extreme R/K-rich toxic sequences improved solubility profiles

3. Design quality PepMLM = “sampling plausible binders” moPPIt = “engineering binders with constraints”

4. Evaluation before clinical progression Before advancing any peptide: Structural stability check AlphaFold3 ipTM consistency across replicates Specificity binding must localize near A4V region or functional interface ADMET-like properties (PeptiVerse) low hemolysis risk high solubility balanced net charge Robustness mutation tolerance of binding site off-target binding screening

Final conclusion

moPPIt peptides outperform PepMLM peptides in:

binding localization therapeutic balance design controllability

However, PepMLM remains useful for diverse candidate exploration, while moPPIt is superior for clinical-grade optimization pipelines.

Part B: BRD4 Drug Discovery Platform Tutorial (Gabriele)

(View Full Screen)

Part C: Final Project: L-Protein Mutants

Final Proposal of Group Project