Genomic perplexity and the evolution of context-dependent function.

Area of Science:

• Genomics
• Evolutionary Biology
• Computational Biology

Background:

• Traditional genetics assumes fixed gene functions, but pangenomics and GWAS show context-dependent organismal functions.
• Genomic context significantly influences gene and organismal phenotypes, even for core genes.

Purpose of the Study:

• To propose a novel framework viewing genomes as probability distributions over functional outcomes, analogous to large language models (LLMs).
• To introduce and define 'genomic perplexity' as an information-theoretic measure of genetic element incompatibility within a genomic context.
• To provide a testable framework for predicting gene integration potential and understanding evolutionary processes.

Main Methods:

• Conceptual analogy between genomic epistasis and LLM attention mechanisms.
• Introduction of information-theoretic concept 'genomic perplexity'.
• Demonstration of perplexity as a metric for fitness costs in horizontal gene transfer (HGT) and introgression.

Main Results:

• Genomes, like LLMs, encode probability distributions of function, not fixed functions.
• Genomic epistasis is analogous to attention mechanisms, where context weights genetic element influence.
• Genomic perplexity quantifies the incompatibility of genetic elements, explaining fitness costs in gene flow.

Conclusions:

• A probabilistic, LLM-inspired framework reframes our understanding of gene function and genomic interactions.
• Genomic perplexity offers a quantifiable metric for gene-environment interactions and evolutionary integration.
• This perspective advances synthetic biology, evolutionary modeling, and our understanding of genomic adaptation.