joint SNP calling for individuals with very different coverage

[bmartinez]

We have resequenced the genome of 40 individuals from two populations of each of two species to do population genomic analyses. Different individuals were sequenced at very different depth (60x, 28x, 5x). Our idea was to perform a joint SNP calling (per species) directly, but we have been alerted that mixing samples with different coverage might results in biases and recommended to equalize coverage beforehand by downsampling higher coverage samples to 5x. Intuitivelly, we tend to think that inclusion of high coverage samples should increase overall reliability of variant calling and genotyping and do not understand the type of biases this might introduce. Could anyone please provide any input? What would be the best way of combining samples resequenced at different depths in variant calling and genotyping?

Thanks a lot in advance,

BMartinez