Bldg. 38A, new NCBI library

Tuesday, March 30, 2004, 11 AM

 

What's important in life?

Connections between measures of genome evolution and biological quantities

 

Eugene V. Koonin, NCBI/NLM/NIH


In the last few years, we witnessed the advent of multiple, complete genome
sequences and, in parallel, of genome-wide data on biological properties of
gene ensembles, such as knockout effect, expression levels, protein-protein
interactions, and more. Inevitably, numerous attempts were made by several
groups, including ours, to examine connections between these properties and
quantitative measures of gene evolution. The question asked are quite
fundamental and interesting such as: What determines the effect of the
knockout of a given gene on the phenotype (in particular, is it essential or
not) and the rate of a gene's evolution? And how are the phenotypic effect
and evolutionary rate connected? And more questions like that. The best
succinct description of the results of the genome-wide studies addressing
these questions seems to be: it is a mess. Many tantalizing correlations
have been detected and made into a big deal (or not). For example, a
positive correlation was detected between the tendency of a gene to be lost
during evolution and sequence evolution rate, and negative correlations were
noticed between each of the above measures of evolutionary variability and
phenotypic effect or expression level. Simply put, genes associated with a
major phenotypic effect tend to be highly expressed, form many
protein-protein interactions and evolve slowly, however that is measured.
These connections seem to meet the expectations based on common sense. What
is quite unexpected, however, is that, although many of the observed
correlations are statistically highly significant (mostly thanks to the
large number of data points), they are typically rather weak. Linear
correlation coefficients in the range of 0.1-0.2 explaining only a small
fraction of the scatter in the data are typical. This makes detecting real
causal relationships a serious challenge. In this talk, I will give an
overview of the reported genomic correlations; present in some detail a few
examples explored in our group; describe preliminary attempts to separate
the wheat from the chaff using multivariate statistical analysis; and  to
outline a generalized approach, also very preliminary. I will introduce the
notion of the "social status" (or, simply, "importance") of a gene in the
genome-wide community and classify the variables with respect to their
positive or negative correlation with this status. I will also argue that,
despite the complexity of the emerging web of relationships, there is a
clearly discernible pattern, and deviations from it point to either problems
with the data or biologically interesting phenomena. 

Joint work with Yuri Wolf, with thanks to Liran Carmel, I. King Jordan,
Fyodor Kondrashov, Dmitry Krylov, and Igor Rogozin.