Leonid Meyerguz

CS 789 THEORY SEMINAR [home]

Speaker:  Leonid Meyerguz    
Affiliation: Cornell University    
Date: Monday, May 19, 2003
Title: On the Temperature of Evolution

Abstract:

It is well known that there is considerable redundancy in the sequence-to-structure map for proteins; large numbers of homologous

proteins cluster in the neighborhood of characteristic three-dimensional shapes. It is an open question, however, whether these clusters

form isolated islands in sequence space, or whether the space of functional proteins is highly connected, allowing for gradual

transformations from one class of proteins to another by a sequence of small mutational steps.  While evidence for extensive connectivity

has been found in simplified models of protein structure, it has been extremely difficult to gain insight into this issue on the set of all

amino acid sequences with a reasonably accurate fitness function, due to the enormous size and complexity of the search space. Here

we provide a novel source of evidence for the connectivity of sequence space, using a randomized counting algorithm and techniques

adapted from statistical mechanics. We compute a notion of evolutionary temperature for protein sequences, and find that

native sequences of more than 200 amino acids have roughly constant temperature. This apparent equilibrium in sequence space

provides a 'fingerprint' of global connectivity, suggesting either that there has been evolutionary migration between distinct protein

families over time, or that a uniform evolutionary mechanism is operating in each `island' to preserve this temperature equilibrium.

Joint work with Ron Elber and Jon Kleinberg.