Machine-Learning Guided Protein Engineering

Question 1

Explain “protein landscape” given example of protein with 12 aas.

Answer 1

• Protein sequence space represents all possible sequences for a protein or gene.
• Sequence space has one dimension per amino acid in the sequence -> highly dimensional spaces
• DE visualized as a series of steps within a 3D fitness landscape
• 20^n unique variations of the protein, n = number of amino acids in the chain, 20^12 = 4x10^15 variants

Question 2

Explain DML.

Answer 2

• Predicts how multiple mutations in RBD of CoV2 will impact RBS function (ACE2 binding or Ab escape)
• Combinatorial mutation libraries are generated via yeast display
• ACE2 binding population is identified via flow cytometry
• Categorize variants into population that escapes Ab and one that doesn’t
• Deep sequence the selected variants
• Feed the function and sequence info to the model
• Generates data on positions and function landscapes
• Model predicts probability of binding to ACE2 (predicts escape, synthetic and natural lineages, possible future variants)