Week 5 Lab: Protein Design Part II

Lab Option 1: L-Protein Mutational Analysis

Experimental Validation & DMS

This lab provides a Deep Mutational Scanning (DMS) style validation of the L-Protein. By cross-referencing experimental lysis results—where a score of 1 indicates functional lysis and 0 indicates non-functional—with the Log-Likelihood Ratio (LLR) Heatmap generated via ESM2, we can assess the predictive power of protein language models.

Key Findings

  • Correlation: The agreement between the experimental data and the LLR Heatmap is remarkably high, approximately 90-95%.
  • Predictive Accuracy: The heatmap serves as a highly reliable predictor of whether a specific mutation will allow the protein to retain its ability to lyse bacterial cells.
  • Challenges: Designing mutants with high computational confidence remains difficult, highlighting the current limitations of some structure-based models.
ESM2 LLR Mutational Heatmap ESM2 LLR Mutational Heatmap

Targeted Mutation Strategy

To identify promising variants, I cross-referenced the ESM2 scores with experimental lab data, specifically looking for residues that are not strictly conserved (via pBLAST) and show positive mutational effects.

Selection Criteria:

  1. Soluble Region (N-tail): Targeted to assess how surface-exposed changes affect function.
  2. Transmembrane Region: Targeted to test the hypothesis that the L-protein assembles to perforate the bacterial membrane.
  3. Combined Effects: Testing synergistic effects of multiple “positive-score” mutations.

Proposed Mutants

MutantSubstitutionLocationIndexSequence Snippet
1P –> QSoluble (N-tail)6METRFQQQSQQTPASTNRRRPFKHEDYPCRRNQRSST...
2C –> SSoluble (N-tail)29METRFPQQSQQTPASTNRRRPFKHEDYPSRRNQRSST...
3S –> LTransmembrane49...LYVLIFLAIFLLKFTNQLLLSLLEAVIRTVTTL...
4K –> LTransmembrane50...LYVLIFLAIFLSLFTNQLLLSLLEAVIRTVTTL...
5C, K –> S, LCombined29 & 50...PSRRNQRSSTLYVLIFLAIFLSLFTNQLLLSL...

Structural Hypothesis: Multimeric Assembly

A running hypothesis for L-protein function is that it assembles into a multimeric complex to create perforations in the bacterial membrane. To investigate this, I utilized AF2_Multimer to generate a predicted multimeric assembly.

By modeling these specific mutations (particularly those in the transmembrane region like Mutant 3 and 4) in a multimeric context, we can observe if the substitutions stabilize the pore-forming structure or increase the efficiency of membrane lysis.

Gemini AI was consulted for data organization and Markdown formatting