Week 5 HW: PROTEIN DESIGN PART II

HTGAA 2026 — Week 5 Homework

PROTEIN DESIGN PART II

Part 1: Generating Peptides with PepMLM

SOD1 A4V Sequence (Human):

MATVAVCVLKGDGPVQGIINFEQKESNGPVKVWGSIKGLTEGLHGFHVHEFGDNTAGCTSAGPHFNPLSRKHGGPKDEERHVGDLGNVTADKDGVADVSIEDSVISLSGDHCIIGRTLVVHEKADDLGKGGNEESTKTGNAGSRLACGVIGIAQ

Generated Peptides: We obtained approximately 4 peptides plus one control peptide, each showing a perplexity score. The sequence modification made was Alanine to Valine substitution at position 4.

• Pép-1: WVRFLPYRKGGL (Perplexity: 12.4)
• Pép-2: KLVFAVGPSTRR (Perplexity: 14.8)
• Pép-3: RRLYWAIPRGGF (Perplexity: 11.2)
• Pép-4: VVFQRLSSTGRR (Perplexity: 15.1)
• Control: FLYRWLPSRRGG (Known peptide)

The low perplexity observed in Pép-3 suggests that the model has high confidence that this sequence “fits” in the evolutionary and structural context of the mutated SOD1. Lower perplexity indicates greater model confidence in the sequence’s contextual appropriateness.

Part 2: Evaluation with AlphaFold3 (AF3)

When these sequences were loaded into AF3 along with SOD1 A4V, the following results were obtained:

Pép-3 (RRLYWAIPRGGF):

• iPTM (interaction Predicted TM-score): 0.78
• Location: Near the N-terminal region
• Function: Appears to stabilize the beta sheet destabilized by the A4V mutation
• Strategic positioning: Close to the mutation site (position 4)

Control (FLYRWLPSRRGG):

• iPTM: 0.82
• Location: Typically binds at the dimer interface
• Interaction pattern: High-confidence dimer interface binding

Both iPTM values exceed 0.7, indicating high confidence in the interactions. iPTM scores above 0.7 are generally considered excellent predictors of binding reliability. In the context of our final project, the MeR peptide binds near the metal-binding site (such as the Cys-Cys site in MerA), which is fundamental for preventing mercury-induced protein unfolding before reduction occurs.

Part 3: Property Assessment in PeptiVerse

Although the control peptide exhibits the highest affinity (8.5), its low solubility and high hemolytic potential make it a poor therapeutic candidate. Therefore, Pép-3 is selected as the optimal candidate, achieving an excellent balance between affinity and favorable physicochemical properties. Its positive charge (+3) and high solubility facilitate interaction with acidic regions of SOD1 (or the MerA enzyme used in our final project), promoting proper protein folding in the bacterial cytoplasm during bioremediation processes.

Part 4: Optimization with moPPIt

In moPPIt, we do not simply generate random sequences; instead, we guide the design process toward specific objectives which are:

1. Design a peptide that specifically binds to residues 4-10 (A4V mutation site)
2. Target zinc-binding residues (positions 80, 71, 63) However it is important to value these things:

• PepMLM: Proposes “natural” sequences based on language model training
• moPPIt: Generates mathematically optimized sequences specifically designed for enhanced affinity This two-stage approach combines evolutionary relevance (PepMLM) with computational optimization (moPPIt) to create peptides with superior binding characteristics.

Part 5: Application in MerR for Bioremediation (Final Project)

My final project is a -> mercury biosensor design:

Using the MerR protein as a heavy metal sensor (particularly for mercury), we employ moPPIt to design peptides that mimic the Hg(II) binding site. This approach allows us to:

1. Create more sensitive biosensors by engineering peptide sequences that specifically recognize mercury ions
2. Improve binding specificity through optimized interactions at the metal-binding interface
3. Enhance bioremediation efficiency by tailoring peptide-protein interactions to target specific mercury-binding mechanisms