Week 5 HW: Protein Design Part II

SOD1 Binder Peptide Design (From Pranam)

Generate Binders with PepMLM

PepMLM Colab File

SOD1 sequence: MATKAVCVLKGDGPVQGIINFEQKESNGPVKVWGSIKGLTEGLHGFHVHEFGDNTAGCTSAGPHFNPLSRKHGGPKDEERHVGDLGNVTADKDGVADVSIEDSVISLSGDHCIIGRTLVVHEKADDLGKGGNEESTKTGNAGSRLACGVIGIAQ

Sequence with A4V mutation: MATKVVCVLKGDGPVQGIINFEQKESNGPVKVWGSIKGLTEGLHGFHVHEFGDNTAGCTSAGPHFNPLSRKHGGPKDEERHVGDLGNVTADKDGVADVSIEDSVISLSGDHCIIGRTLVVHEKADDLGKGGNEESTKTGNAGSRLACGVIGIAQ

Binders, perplexity score:

  • WRVPPAALRHKE, 22.653588
  • HRSPPVAAEHWK, 19.512332
  • WRYYPVAAAWKK, 11.081843
  • WRYYVAALRHGK, 15.691672
  • known SOD1 binder: FLYRWLPSRRGG, 20.635231

Evaluate Binders with AlphaFold3

BinderipTM ScoreBinding Location
WRVPPAALRHKE0.39near β-barrel and globular part
HRSPPVAAEHWK0.26near β-barrel, across the seam and onto globular part
WRYYPVAAAWKK0.28near β-barrel, across the seam
WRYYVAALRHGK0.30near β-barrel, across the seam
known: FLYRWLPSRRGG20.6352310.31

The ipTM score is highest for the first binder, but none of the binders bind in a similar location compared to the known binder. All of the predicted binders bound a similar part of the protein that was very different from where the known binder is predicted to bind.

Evaluate Properties of Generated Peptides in the PeptiVerse

BinderBinding Affinity (pKd/pKi)Solubility ProbabilityHemolysis ProbabilityNet ChargeMolecular Weight (Da)
WRVPPAALRHKEWeak, 5.114Soluble, 1Non-hemolytic, 0.0201.851450.7
HRSPPVAAEHWKWeak, 4.576Soluble, 1Non-hemolytic, 0.0140.941414.6
WRYYPVAAAWKKWeak, 6.124Soluble, 1Non-hemolytic, 0.0212.761538.8
WRYYVAALRHGKWeak, 6.474Soluble, 1Non-hemolytic, 0.0232.841519.8
known: FLYRWLPSRRGGWeak, 5.968Soluble, 1Non-hemolytic, 0.0472.761507.7

Binding affinity is weak for all peptides. There does not seem to be a strong correlation between ipTM score and binding affinity. From PeptiVerse results, it appears that all peptides are soluble and non-hemolytic. The ones with stronger charges have a higher binding affinity.

I’m choosing to continue with the 4th generated peptide, WRYYVAALRHGK, because it has the highest binding affinity from PeptiVerse and it has a ipTM score similar to the known peptide from AlphaFold.

Generate Optimized Peptides with moPPIt

Motif positions used: 95-100, 85-90

Peptide generated: PKHCLQRLLSKH

  • ipTM score: 0.45
  • Binding location: end of the β-barrel that is closer to the termini
  • Binding affinity (pKd/pKi): weak, 6.408
  • Solubility probability: soluble, 1
  • Hemolysis probability: non-hemolytic, 0.057
  • Net charge: 3.12
  • Molecular weight (Da): 1459.8

This peptide has one of the strongest binding affinities according to PeptiVerse along with the highest ipTM score from AlphaFold. Like the other peptides, it is soluble and non-hemolytic. Consistent with previous observations, it has the strongest charge, which could be why it binds mutated SOD1 relatively strongly.

Final Project: L-Protein Mutants

L-protein sequence: METRFPQQSQQTPASTNRRRPFKHEDYPCRRQQRSSTLYVLIFLAIFLSKFTNQLLLSLLEAVIRTVTTLQQLLT
Last 35 residues are the transmembrane domain

DnaJ sequence: MAKQDYYEILGVSKTAEEREIRKAYKRLAMKYHPDRNQGDKEAEAKFKEIKEAYEVLTDSQKRAAYDQYGHAAFEQGGMGGGGFGGGADFSDIFGDVFGDIFGGGRGRQRAARGADLRYNMELTLEEAVRGVTKEIRIPTLEECDVCHGSGAKPGTQPQTCPTCHGSGQVQMRQGFFAVQQTCPHCQGRGTLIKDPCNKCHGHGRVERSKTLSVKIPAGVDTGDRIRLAGEGEAGEHGAPAGDLYVQVQVKQHPIFEREGNNLYCEVPINFAMAALGGEIEVPTLDGRVKLKVPGETQTGKLFRMRGKGVKSVRGGAQGDLLCRVVVETPVGLNERQKQLLQELQESFGGPTGEHNSPRSKSFFDGVKKFFDDLTR

Mutagenesis

L-protein mutation likeihood heatmap: heatmap heatmap

Experimental data of L-protein mutations

For the most part, experiments where lysis still occured correlates to a yellowish green point on the heatmap, suggesting that the heatmap is somewhat reliable.

Proposed mutations:

MutationDomainReasoningL-protein Multimer
R19SSolubleThis amino acid is shown to change in BLASTp results, experimental evidence suggests this results in a lytic protein. The heatmap score does not appear to be strongly negative.R19S R19S Multimer forms that appears circular and pore-like. However, the ipTM score is 0.14, which is low, suggesting low confidence in this assembly of a multimer.
R31ISolubleThis amino acid is shown to change in BLASTp results, experimental evidence suggests this results in a lytic protein. The heatmap score does not appear to be strongly negative.R31I R31I Multimer forms that appears circular and pore-like. However, the ipTM score is 0.16, which is low, suggesting low confidence in this assembly of a multimer.
H24LSolubleThis amino acid is shown to change in BLASTp results, and the heatmap shows a very positive score for this mutation. There is no experimental data, so it is possible this mutation would result in a functional protein, even though other amino acid substitutions at this position have not, especially since histidine and lysine are both positively charged amino acids.H24L H24L Multimer forms that appears circular and pore-like. However, the ipTM score is 0.15, which is low, suggesting low confidence in this assembly of a multimer.
T69STransmembraneThis amino acid is shown to change in BLASTp results, and the heatmap shows a score near 0, so it’s not strongly negative. However, experimental data suggests this protein does not result in lysis. This makes sense because if the transmembrane domain is mutated, the L-protein will not be able to oligomerize in the membrane and create pores.T69S T69S Multimer forms that appears circular and pore-like. However, the ipTM score is 0.15, which is low, suggesting low confidence in this assembly of a multimer.
N53TTransmembraneThis amino acid is shown to change in BLASTp results, and the heatmap shows a very positive score for this mutation. There is no experimental data, so it is possible this mutation would result in a functional protein, even though other amino acid substitutions at this position have not. Also, this mutation is the only polar-to-polar mutation that was not experimentally tested.N53T N53T Multimer forms but appears asymmetrial. ipTM score is 0.16, which is low, suggesting low confidence in this assembly of a multimer.

Mutagenesis using Af2-Multimer

Proposed mutations:

MutationDomainReasoningL-protein-DnaJ Multimer
R19SSolubleThis amino acid is shown to change in BLASTp results, experimental evidence suggests this results in a lytic protein. The heatmap score does not appear to be strongly negative.R19S_DnaJ R19S_DnaJ The L-protein and DnaJ appear to make contact. ipTM is 0.17, which is low, suggesting low confidence in this interaction.
R31ISolubleThis amino acid is shown to change in BLASTp results, experimental evidence suggests this results in a lytic protein. The heatmap score does not appear to be strongly negative.R31I_DnaJ R31I_DnaJ The L-protein and DnaJ appear to make contact. ipTM is 0.16, which is low, suggesting low confidence in this interaction.
H24LSolubleThis amino acid is shown to change in BLASTp results, and the heatmap shows a very positive score for this mutation. There is no experimental data, so it is possible this mutation would result in a functional protein, even though other amino acid substitutions at this position have not, especially since histidine and lysine are both positively charged amino acids.H24L_DnaJ H24L_DnaJ The L-protein and DnaJ appear to make contact. ipTM is 0.16, which is low, suggesting low confidence in this interaction.
F5SSolubleThis particular substitution exists in BLASTp results, and the heatmap shows a very positive score. There is no experimental data, so it is possible this mutation would result in a functional protein.F5S_DnaJ F5S_DnaJ The L-protein and DnaJ appear to make contact. ipTM is 0.16, which is low, suggesting low confidence in this interaction.
A14SSolubleThis amino acid is shown to change in BLASTp results, and the heat map shows a positive score. There is no experimental data, so it is possible this mutation would result in a functional protein.A14S_DnaJ A14S_DnaJ The L-protein and DnaJ appear to make contact ipTM is 0.16, which is low, suggesting low confidence in this interaction.