Abstract
Recombination results in the reciprocal exchange of genetic information occurring in meiosis which increases genetic variation by producing new haplotypes. Recombination rates are heterogeneous between species and also along different genomic regions. Large fractions of recombination events are often concentrated on short segments known as re-combination hotspots. In this work we statistically inferred heterogeneous recombination rates by using relevant summary statistics as explanatory variables in a regression model. We used for this purpose a frequentist segmentation algorithm with type I error control to estimate the variation in local recombination rates. Under various simulation setups we have obtained very fast and accurate estimates. We also show an example of an inference on a 103kb region of the human genome and compare our inferred historical- and population-specific hotspots with results from experimental data (sperm-typing and double strand break maps). For the analyzed region, our method shows a good congruence between historical and experimental hotspots, except for one hotspot hypothesized to be population-specific. This method is implemented in the R-package LDJump, which is available from https://github.com/PhHermann/LDJump.