UC Berkeley researchers use algorithm to explain sex

Related Posts

A paper published Monday and co-authored by UC Berkeley researchers uses an algorithm to explain the diversity sexual recombination yields.

The paper, published in the online Early Edition of the Proceedings of the National Academy of Sciences, was a collaboration between UC Berkeley computer scientists and researchers in various other scientific and mathematical fields. It takes a well-known algorithm, often used in computer science, and applies it to the field of evolutionary biology.

Adi Livnat, an author of the paper and an assistant professor of biological sciences at Virginia Tech, said the findings bring together three theoretical fields: algorithms, evolutionary theory and game theory, the study of strategic decision-making.

One question of the study is that of the mystery of genetic variation, according to Christos Papadimitriou, a UC Berkeley professor of electrical engineering and computer sciences. If the point of evolution is to select the best genes, he asked, then why is there so much genetic variation?

“From the viewpoint of evolution and biology, the question is, ‘what is the role of sexual recombination?'” said Umesh Vazirani, a UC Berkeley professor of electrical engineering and computer sciences. “From the viewpoint of computer science, the question is, ‘what is the algorithm for evolution?’”

Vazirani said the paper’s findings unite these questions and suggest sexual recombination allows for a repeated “game” for genes to play that uses an algorithm called the multiplicative weight updates algorithm.

The algorithm maximizes a tradeoff between cumulative performance and entropy. In other words, Vazirani said the algorithm maximizes the tradeoff between going all in on a successful genetic trait and hedging bets among a wider variety of genes.

For example, in economics, the algorithm can be utilized for adjusting investment in an array of stocks based on stock performance, Papadimitriou said. Hedging one’s bets in such a way creates a similar effect as would result from knowing what the best-performing stocks were in advance.

“It’s the same (in evolutionary biology),” he said. “The players who participate in the game are the genes, and they look at how the various alleles performed in the previous generation and boost the good performers and decrease a little the bad performers.”

The paper focuses on the regime of weak selection in evolution, in which one trait is only slightly more advantageous compared to another. In this way, selection forces are not sudden, only offering gentle advantages that amount to slow improvement over time, Papadimitriou said.

Vazirani said applying this algorithm to evolutionary biology — thus connecting these multiple fields — is exciting, although these findings have just scratched the surface of what can be explored. According to the paper, one new research direction would be to introduce genes with mutations to the current analysis.

Contact Nico Correia at [email protected] and follow him on Twitter @nicolocorreia.