Historically, it has been postulated that populations carry genetic variants with deleterious effects as segregating variation. Recent advances in DNA resequencing technology allow a new view on this topic, providing a way to identify segregating molecular sequence variants that alter sites that show constrained evolution. Targeted identification and removal of such segregating variants has been proposed as a novel path for improving agronomic performance in plant breeding programs. In this dissertation, I present a brief review of deleterious variation, and why it may be important to plant breeding. Then, I present a tool for classifying single nucleotide polymorphisms into functional classes using public sequence databases. Next, I present a survey of potentially deleterious alleles segregating in two crop species, and a software package that implements a likelihood ratio test for sequence constraint and deleterious prediction. Finally, I present an examination of the contribution of potentially deleterious alleles to yield in a barley breeding experimental population. While potentially deleterious alleles themselves do not explain a larger proportion of phenotypic variation than non- deleterious alleles, genomic prediction of phenotypes may be improved by examining informative subsets of sequence variants, rather than with genomewide markers.