ABSTRACT
In the last decades the amazing reduction in DNA sequencing costs, resulted in the application of human whole exome and/or genome sequencing in the clinical practice. In the context of rare disease molecular diagnosis, the goal is to determine which of the hundreds of thousands to millions of variants found in each patient, is responsible for the development of the observed disease. In other words, the key issue is to be able to determine which variant is pathogenic -the needle- among the haystack of benign polymorphisms.
Bioinformatics techniques significantly contribute to this task, particularly when applied to the benign/pathogenic prediction and ranking of novel variants. This chapter describes and reviews different strategies and algorithms for variant effect prediction, comparing their foundations, applicability and adoption by the scientific and medical community. In addition, relevant issues related to algorithm development are described, pointing to potential pitfalls and opportunities. Finally, the authors describe their own contribution, which focuses not directly on prediction but on understanding how and why a variant is pathogenic in the context of the corresponding protein structure-function relationships.
