Locality-sensitive hashing enables efficient and scalable signal classification in high-throughput mass spectrometry raw data

• DOI: 10.1186/s12859-022-04833-5
• Publication Date: 2022-07-20T14:04:07Z
• Authors: Konstantin Bob; David Teschner; Thomas Kemmer; David Gomez-Zepeda; Stefan Tenzer; Bertil Schmidt; Andreas Hildebrandt
• Keywords: Locality-sensitive hashing

Mass spectrometry is an important experimental technique in the field of proteomics. However, analysis certain mass data faces a combination two challenges: first, even single experiment produces large amount multi-dimensional raw and, second, signals interest are not peaks but patterns that span along different dimensions. The rapidly growing increases demand for scalable solutions. Furthermore, existing approaches signal detection usually rely on strong assumptions concerning properties.In this study, it shown locality-sensitive hashing enables classification at scale. Through appropriate choice algorithm parameters possible to balance false-positive and false-negative rates. On synthetic data, superior performance compared intensity thresholding approach was achieved. Real could be strongly reduced without losing relevant information. Our implementation scaled out up 32 threads supports acceleration by GPUs.Locality-sensitive desirable data.Generated code available https://github.com/hildebrandtlab/mzBucket . Raw https://zenodo.org/record/5036526