Week 5 HW: Protein Design Part II

Part A: SOD1 Binder Peptide Design (From Pranam)

Part 1: Generate Binders with PepMLM

This is the human SOD1 sequence containing the A4V mutation: MATKVVCVLKGDGPVQGIINFEQKESNGPVKVWGSIKGLTEGLHGFHVHEFGDNTAGCTSAGPHFNPLSRKHGGPKDEERHVGDLGNVTADKDGVADVSIEDSVISLSGDHCIIGRTLVVHEKADDLGKGGNEESTKTGNAGSRLACGVIGIAQ

	Binder	Pseudo Perplexity
0	WHYYATGARWGE	16.929015
1	WRYGAVALELKK	12.714672
2	WRSPAAAARWWK	9.155765
3	WRYPATAAALKX	4.843841
4	FLYRWLPSRRGG	N/A

Info

The table generated by PepMLM detailing possible peptides to bind to mutant SOD1 along with their pseudo perplexity scores. Peptide 4 is an already known SOD1-binding peptide.

Part 2: Evaluate Binders with AlphaFold3

	Binder	ipTM	Peptide localizes near SOD1 N-terminus?	Engages β-barrel region?	Approaches dimer interface?	Surface-bound/partially buried peptide?
0	WHYYATGARWGE	0.32	No	No	Yes	Yes
1	WRYGAVALELKK	0.18	No	No	Yes	Yes
2	WRSPAAAARWWK	0.42	No	Yes	No	Yes
3	WRYPATAAALKA	0.46	No	Yes	No	Yes
4	FLYRWLPSRRGG	0.29	No	No	Yes	Yes

Info

Table containing results from AlphaFold3 generations

Peptide generations 0, 1, and 4 contain ipTM scores lower than 0.4, demonstrating no meaningful binding between the peptide and the A4V mutated SOD1 protein. However, compared to the known binder 4, peptides 0, 2, and 3 have greater ipTM values. Peptides 2 and 3 contain ipTM scores significantly greater than the known binder 4 at 0.42 and 0.46, respectively.

Part 3: Evaluate Properties of Generated Peptides in the PeptiVerse

The results from PeptiVerse of the 5 peptides proved to be mostly consistent with the ipTM values in AlphaFold3. Peptides 0, 1, and 4 resulted in weak binding which is consistent with their low ipTM scores in AlphaFold3. However, peptide 2 also resulted in weak binding in PeptiVerse which was unexpected as it had an ipTM score of 0.42 in AlphaFold3. Peptide 3 resulted in medium binding which is consistent with its ipTM score of 0.46. All five peptides were soluable and non-hemolytic so they all had sufficient therapeutic properties. Only peptide 4 balanced predicted binding and therapeutic properties well.

Part 4: Generate Optimized Peptides with moPPIt

Peptide Sequence	Hemolysis ↓	Solubility ↑	Affinity ↑	Motif Match ↑
GGKKEYYYSRYP	0.9586	0.9167	7.21	0.1572
EKQYTCDTSTKM	0.9675	0.9167	6.18	0.8416
KKTTGYGECSYN	0.9639	1.0000	5.85	0.8290
GTYTCETTYTQW	0.9728	0.9167	6.68	0.8369

Info

Table of results from the moPPIt-v3 Colab using the A4V mutated SOD1 protein.

The peptides generated by the moPPIt-v3 Colab generated stronger binding and motif match results than the peptides generated by PepMLM. In addition, both groups of peptides had high solubility. However, the peptides generated by moPPIt-v3 demonstrated very high hemolysis results, insinuating high risks for red blood cell damage. I would evaluate these peptides to be unready for clinical studies since their hemolytic values are too high. These peptides would be too dangerous to use in therapeutic applications.

Part C: Final Project: L-Protein Mutants

I worked with Jason Ross, Xavier-Lewis Palmer, and Nana Agyei Afrane-Asare to generate mutated proteins to improve the stability of the L-Protein. Our results can be found in this Google Doc.