Part A:

PepMLM for SOD1 mutated at 4 A2V produced these peptides:

After running AlphaFold-Multimer on the 5 peptides with SOD1 mutant: MATKVVCVLKGDGPVQGIINFEQKESNGPVKVWGSIKGLTEGLHGFHVHEFGDNTAGCTSAGPHFNPLSRKHGGPKDEERHVGDLGNVTADKDGVADVSIEDSVISLSGDHCIIGRTLVVHEKADDLGKGGNEESTKTGNAGSRLACGVIGIAQ

All prediction gave low ipTM which suggest poor model prediction. This could be since we are missing a specific parameter in the model to give better predictions but I was unable to figure it out.

It is fair to assume that peptides with bulkier amino acids. A4V mutation in SOD1 will likely lead to aggregation and misfolding of protein which will make prediction harder.

Part B:

Option2: Random N1 (can set) point mutations function was made in google colab

https://colab.research.google.com/drive/1h6oKTkeYSckwDhqbjx4oozQdJp3aXGMe?usp=drive_link

Example output of the code for the L protein based on the sheet provided for point mutations looking for random options including only entries that gives a protein and lysis (https://docs.google.com/spreadsheets/d/11WzDDNkQDEiqbUSGV0ZCqITGctyNFpD7xnPlhsj2BhE/edit?gid=0#gid=0) gave (N1=2):

Again AlphaFold-Multimer gave poor ipTM for all peptides under 0.3 suggesting a systemic problem in my execution of the tool. Having said that additional filters which can be used to pick peptides to test can include: (1) validate all other viral components translation is not interrupted

(2) Remove peptides with mutations in core regions affecting lysis

(3) Avoid larger amino acid mutations (from smaller to larger) or change in polarity