liveBook · Manning

Epistasis

As explained in Build AI Drug Discovery Pipelines

Epistasis is the dependence of a mutation's effect on the presence of other mutations in the same sequence.

Epistasis occurs when the effect of a mutation depends on other mutations in the same sequence. In protein design, mutations that look promising individually may interact unfavorably when combined. section 12.3.5

ESM-2 can help flag potentially problematic combinations using pseudo-perplexity (PPPL), a measure of whole-sequence plausibility. It masks each amino-acid position in turn and scores the actual residue. Lower PPPL means the sequence is more typical of the model’s training distribution; it does not guarantee stability, folding, binding, expression, or biological function. section 12.3.5

Combining antibody mutations

Suppose three mutations in an antibody framework each have positive individual log-likelihood ratios. Compare the PPPL of the variant containing all three with that of the wild type. A higher variant PPPL suggests the combination may create an implausible sequence pattern, warranting more conservative testing. This comparison alone does not prove an epistatic interaction. section 12.3.5

How Build AI Drug Discovery Pipelines covers it

How deep it goes
Explains a model-based plausibility check for mutation combinations but stops short of establishing biological effects.
Examples
Illustrates three individually promising antibody framework mutations evaluated together against the wild-type sequence.
What you can do after
Evaluate combined variants using ESM-2 pseudo-perplexity while recognizing that sequence plausibility does not establish biological function.
How it connects
Connects epistasis with protein mutations, log-likelihood ratios, and pseudo-perplexity.

Also explained in

Compare how 2 books explain epistasis →

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Build AI Drug Discovery Pipelines. The section links point to the parts of the book it draws on.