Understanding the effects that non-synonymous single nucleotide polymorphisms have on the structures of the gene products, the proteins, is important in identifying the origins of complex diseases. A method based on environment-specific amino acid substitutions observed within homologous protein families with known 3D structures was used to predict changes in stability caused by mutations. In the task of predicting only the sign of stability change, our method performs comparably or better to other published methods with an accuracy of 71%. The method was applied to a set of disease associated and non-disease associated mutations and was shown to distinguish the two sets in terms of protein stability. Our method may therefore have application in correlating SNPs with diseases caused by protein instability.