The fine-scale landscape of meiotic gene conversion. A. L. Williams1, J. Blangero2, M. Przeworski1 1) Biological Sciences Department, Columbia University, New York, NY; 2) Texas Biomedical Research Institute, San Antonio, TX.

   Meiotic recombination is essential to the proper alignment and segregation of chromosomes, and produces a haploid genome that is a mosaic of the two parental chromosomes. Among possible resolutions of recombination are crossovers, in which homologous chromosomes reciprocally exchange material, and non-crossover gene conversion events, in which short segments are copied from one homolog to the other across 50-1,000 bp. Despite dramatic progress in our understanding of crossover resolutions of recombination in mammals over the past decade, little is known about gene conversion events.
   We present an analysis of meiotic gene conversion patterns identified using whole genome sequence data from 11 three-generation human pedigrees. We estimate the number of gene conversion events per meiosis and their tract lengths, and compare these to rates of crossing-over while accounting for differences in power. We then examine the location of crossover and non-crossover resolutions and their determinants. Specifically, we focus on an observation that we previously made on the basis of more limited data of complex recombination events in which multiple gene conversion tracts cluster near each other and near crossovers. This unexpected pattern, similar to recombination events reported previously in S. cerevisiae, is inconsistent with canonical models of double strand break repair, and if frequent, would be predicted to lead to a complex correlation structure among variants.
   To characterize the impact of non-crossover resolutions on genomic base composition, we estimate the strength of GC-biased gene conversionthe over-transmission of G or C alleles at GC/AT heterozygous sites. This form of transmission distortion is hypothesized to have a major on impact base composition over evolutionary time, yet there is little direct evidence of its existence. Lastly, we leverage these pedigree sequence data to address the long-standing question of whether recombination events produce de novo mutations by analyzing whether these two classes of events co-localize more than expected from background levels.