Alessandro Petrini1, Marco Mesiti1, Max Schubach2,3, Marco Frasca1, Daniel Danis4, Matteo Re1, Giuliano Grossi1, Luca Cappelletti1, Tiziana Castrignanò5,6, Peter N Robinson4, Giorgio Valentini1,7. 1. Università degli Studi di Milano, AnacletoLab - Dipartimento di Informatica, via Giovanni Celoria 18, 20135 Milano, Italy. 2. Berlin Institute of Health (BIH), Anna-Louisa-Karsch-Straße 2, 10178 Berlin, Germany. 3. Charité - Universitätsmedizin Berlin, Chariteplatz 1, 10117 Berlin, Germany. 4. The Jackson Laboratory for Genomic Medicine, 10 Discovery Drive, Farmington (CT) - 06032, United States of America. 5. CINECA, SCAI SuperComputing Applications and Innovation Department, Via dei Tizii 6, 00185 Roma, Italy. 6. University of Tuscia, Department of Ecological and Biological Sciences (DEB), Largo dell'Università snc, 01100 Viterbo, Italy. 7. CINI National Laboratory in Artificial Intelligence and Intelligent Systems - AIIS, Università di Roma, Via Ariosto 25, 00185 Roma, Italy.
Abstract
BACKGROUND: Several prediction problems in computational biology and genomic medicine are characterized by both big data as well as a high imbalance between examples to be learned, whereby positive examples can represent a tiny minority with respect to negative examples. For instance, deleterious or pathogenic variants are overwhelmed by the sea of neutral variants in the non-coding regions of the genome: thus, the prediction of deleterious variants is a challenging, highly imbalanced classification problem, and classical prediction tools fail to detect the rare pathogenic examples among the huge amount of neutral variants or undergo severe restrictions in managing big genomic data. RESULTS: To overcome these limitations we propose parSMURF, a method that adopts a hyper-ensemble approach and oversampling and undersampling techniques to deal with imbalanced data, and parallel computational techniques to both manage big genomic data and substantially speed up the computation. The synergy between Bayesian optimization techniques and the parallel nature of parSMURF enables efficient and user-friendly automatic tuning of the hyper-parameters of the algorithm, and allows specific learning problems in genomic medicine to be easily fit. Moreover, by using MPI parallel and machine learning ensemble techniques, parSMURF can manage big data by partitioning them across the nodes of a high-performance computing cluster. Results with synthetic data and with single-nucleotide variants associated with Mendelian diseases and with genome-wide association study hits in the non-coding regions of the human genome, involhing millions of examples, show that parSMURF achieves state-of-the-art results and an 80-fold speed-up with respect to the sequential version. CONCLUSIONS: parSMURF is a parallel machine learning tool that can be trained to learn different genomic problems, and its multiple levels of parallelization and high scalability allow us to efficiently fit problems characterized by big and imbalanced genomic data. The C++ OpenMP multi-core version tailored to a single workstation and the C++ MPI/OpenMP hybrid multi-core and multi-node parSMURF version tailored to a High Performance Computing cluster are both available at https://github.com/AnacletoLAB/parSMURF.
BACKGROUND: Several prediction problems in computational biology and genomic medicine are characterized by both big data as well as a high imbalance between examples to be learned, whereby positive examples can represent a tiny minority with respect to negative examples. For instance, deleterious or pathogenic variants are overwhelmed by the sea of neutral variants in the non-coding regions of the genome: thus, the prediction of deleterious variants is a challenging, highly imbalanced classification problem, and classical prediction tools fail to detect the rare pathogenic examples among the huge amount of neutral variants or undergo severe restrictions in managing big genomic data. RESULTS: To overcome these limitations we propose parSMURF, a method that adopts a hyper-ensemble approach and oversampling and undersampling techniques to deal with imbalanced data, and parallel computational techniques to both manage big genomic data and substantially speed up the computation. The synergy between Bayesian optimization techniques and the parallel nature of parSMURF enables efficient and user-friendly automatic tuning of the hyper-parameters of the algorithm, and allows specific learning problems in genomic medicine to be easily fit. Moreover, by using MPI parallel and machine learning ensemble techniques, parSMURF can manage big data by partitioning them across the nodes of a high-performance computing cluster. Results with synthetic data and with single-nucleotide variants associated with Mendelian diseases and with genome-wide association study hits in the non-coding regions of the human genome, involhing millions of examples, show that parSMURF achieves state-of-the-art results and an 80-fold speed-up with respect to the sequential version. CONCLUSIONS: parSMURF is a parallel machine learning tool that can be trained to learn different genomic problems, and its multiple levels of parallelization and high scalability allow us to efficiently fit problems characterized by big and imbalanced genomic data. The C++ OpenMP multi-core version tailored to a single workstation and the C++ MPI/OpenMP hybrid multi-core and multi-node parSMURF version tailored to a High Performance Computing cluster are both available at https://github.com/AnacletoLAB/parSMURF.
Authors: Lucia A Hindorff; Praveen Sethupathy; Heather A Junkins; Erin M Ramos; Jayashri P Mehta; Francis S Collins; Teri A Manolio Journal: Proc Natl Acad Sci U S A Date: 2009-05-27 Impact factor: 11.205
Authors: Marten Jäger; Kai Wang; Sebastian Bauer; Damian Smedley; Peter Krawitz; Peter N Robinson Journal: Hum Mutat Date: 2014-04-09 Impact factor: 4.878
Authors: Clare Turnbull; Richard H Scott; Ellen Thomas; Louise Jones; Nirupa Murugaesu; Freya Boardman Pretty; Dina Halai; Emma Baple; Clare Craig; Angela Hamblin; Shirley Henderson; Christine Patch; Amanda O'Neill; Katherine Smith; Antonio Rueda Martin; Alona Sosinsky; Ellen M McDonagh; Razvan Sultana; Michael Mueller; Damian Smedley; Adam Toms; Lisa Dinh; Tom Fowler; Mark Bale; Tim Hubbard; Augusto Rendon; Sue Hill; Mark J Caulfield Journal: BMJ Date: 2018-04-24
Authors: Dongwon Lee; David U Gorkin; Maggie Baker; Benjamin J Strober; Alessandro L Asoni; Andrew S McCallion; Michael A Beer Journal: Nat Genet Date: 2015-06-15 Impact factor: 38.330
Authors: Daniel Danis; Julius O B Jacobsen; Leigh C Carmody; Michael A Gargano; Julie A McMurry; Ayushi Hegde; Melissa A Haendel; Giorgio Valentini; Damian Smedley; Peter N Robinson Journal: Am J Hum Genet Date: 2021-07-21 Impact factor: 11.025