NeoHeadHunter: an algorithm for the detection, ranking and probabilistic classification of neoepitope candidates

BackgroundThe manufacturing of personalized cancer vaccine requires the accurate identification of neoepitopes, abnormal peptides presented by cancer cells and recognized by the host immune system of the cancer patient. ResultsWe designed and developed NeoHeadHunter, a computational algorithm and pipeline to detect and rank neoepitope candidates. Unlike other algorithms, NeoHeadHunter can estimate the probability that each predicted neoepitope candidate is true positive. To evaluate NeoHeadHunter, we used the Tumor neoantigEn SeLection Alliance (TESLA) data-set derived from the sequencing of nine patients and characterized by 44 experimentally validated positive neoepitopes, a data-set derived from the sequencing of three cancer patients and characterized by eight experimentally validated positive neoepitopes and a manually curated data-set consisting of 64 experimentally validated positive neoepitopes. Our evaluation shows that NeoHeadHunter performs the best compared with other algorithms for both detecting and ranking neoepitope candidates and that NeoHeadHunter can accurately predict such probabilities. ConclusionsNeoHeadHunter can increase the effectiveness of personalized cancer vaccine by sensitively detect, accurately rank and probabilistically classify neoepitope candidates. NeoHeadHunter is released under the APACHE-II license at https://github.com/XuegongLab/neoheadhunter for academic use.