Genetic codes optimized as a traveling salesman problem. Academic Article uri icon

Overview

abstract

  • The Standard Genetic Code (SGC) is robust to mutational errors such that frequently occurring mutations minimally alter the physio-chemistry of amino acids. The apparent correlation between the evolutionary distances among codons and the physio-chemical distances among their cognate amino acids suggests an early co-diversification between the codons and amino acids. Here we formulated the co-minimization of evolutionary distances between codons and physio-chemical distances between amino acids as a Traveling Salesman Problem (TSP) and solved it with a Hopfield neural network. In this unsupervised learning algorithm, macromolecules (e.g., tRNAs and aminoacyl-tRNA synthetases) associating codons with amino acids were considered biological analogs of Hopfield neurons associating "tour cities" with "tour positions". The Hopfield network efficiently yielded an abundance of genetic codes that were more error-minimizing than SGC and could thus be used to design artificial genetic codes. We further argue that as a self-optimization algorithm, the Hopfield neural network provides a model of origin of SGC and other adaptive molecular systems through evolutionary learning.

publication date

  • October 28, 2019

Research

keywords

  • Genetic Code
  • Models, Genetic

Identity

PubMed Central ID

  • PMC6816573

Scopus Document Identifier

  • 85074235498

Digital Object Identifier (DOI)

  • 10.1016/j.tig.2006.08.003

PubMed ID

  • 31658301

Additional Document Info

volume

  • 14

issue

  • 10