A new INDEL detection strategy based on feature extraction

Abstract

Insertions and deletions (indels) are important structural variations in the human genome. Many researchers have created effective detection algorithms; however, the results include a significant amount of false discovery. Increasing the accuracy of indel detection schemes remains a challenge. As coverage has increased, existing methods such as Pindel generate high false-discovery rates, owing to the number of reads. In order to resolve this problem, we present a new indel detection strategy. Our strategy uses related variation features, and combines the Pindel and Adaboost algorithms based on feature extraction. We utilized data from the internationally recognized 1000 Genomes Project for our simulations. The experimental results show that the new strategy effectively increases upon the accuracy of the single indel detection algorithm, and reduces the false discovery rate of the Pindel detection results.