Delineating the effective use of self-supervised learning in single-cell genomics

Guardado en:
Detalles Bibliográficos
Publicado en:Nature Machine Intelligence vol. 7, no. 1 (Jan 2025), p. 68-82
Autor principal: Richter, Till
Otros Autores: Bahrami, Mojtaba, Xia, Yufan, Fischer, David S., Theis, Fabian J.
Publicado:
Nature Publishing Group
Materias:
Acceso en línea:Citation/Abstract
Full Text
Full Text - PDF
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!

MARC

LEADER 00000nab a2200000uu 4500
001 3159907619
003 UK-CbPIL
022 |a 2522-5839 
024 7 |a 10.1038/s42256-024-00934-3  |2 doi 
035 |a 3159907619 
045 2 |b d20250101  |b d20250131 
100 1 |a Richter, Till  |u Department of Computational Health, Institute of Computational Biology, Helmholtz Munich, Munich, Germany (ROR: https://ror.org/00cfam450) (GRID: grid.4567.0) (ISNI: 0000 0004 0483 2525); TUM School of Computation, Information and Technology, Technical University of Munich, Munich, Germany (ROR: https://ror.org/02kkvpp62) (GRID: grid.6936.a) (ISNI: 0000 0001 2322 2966) 
245 1 |a Delineating the effective use of self-supervised learning in single-cell genomics 
260 |b Nature Publishing Group  |c Jan 2025 
513 |a Journal Article 
520 3 |a Self-supervised learning (SSL) has emerged as a powerful method for extracting meaningful representations from vast, unlabelled datasets, transforming computer vision and natural language processing. In single-cell genomics (SCG), representation learning offers insights into the complex biological data, especially with emerging foundation models. However, identifying scenarios in SCG where SSL outperforms traditional learning methods remains a nuanced challenge. Furthermore, selecting the most effective pretext tasks within the SSL framework for SCG is a critical yet unresolved question. Here we address this gap by adapting and benchmarking SSL methods in SCG, including masked autoencoders with multiple masking strategies and contrastive learning methods. Models trained on over 20 million cells were examined across multiple downstream tasks, including cell-type prediction, gene-expression reconstruction, cross-modality prediction and data integration. Our empirical analyses underscore the nuanced role of SSL, namely, in transfer learning scenarios leveraging auxiliary data or analysing unseen datasets. Masked autoencoders excel over contrastive methods in SCG, diverging from computer vision trends. Moreover, our findings reveal the notable capabilities of SSL in zero-shot settings and its potential in cross-modality prediction and data integration. In summary, we study SSL methods in SCG on fully connected networks and benchmark their utility across key representation learning scenarios.Self-supervised learning techniques are powerful assets for enabling deep insights into complex, unlabelled single-cell genomic data. Richter et al. here benchmark the applicability of self-supervised architectures into key downstream representation learning scenarios. 
653 |a Data analysis 
653 |a Datasets 
653 |a Self-supervised learning 
653 |a Computer vision 
653 |a Data integration 
653 |a Empirical analysis 
653 |a Genomics 
653 |a Natural language processing 
653 |a Genes 
653 |a Representations 
653 |a Benchmarks 
653 |a Task complexity 
700 1 |a Bahrami, Mojtaba  |u Department of Computational Health, Institute of Computational Biology, Helmholtz Munich, Munich, Germany (ROR: https://ror.org/00cfam450) (GRID: grid.4567.0) (ISNI: 0000 0004 0483 2525); TUM School of Life Sciences Weihenstephan, Technical University of Munich, Munich, Germany (ROR: https://ror.org/02kkvpp62) (GRID: grid.6936.a) (ISNI: 0000 0001 2322 2966) 
700 1 |a Xia, Yufan  |u TUM School of Computation, Information and Technology, Technical University of Munich, Munich, Germany (ROR: https://ror.org/02kkvpp62) (GRID: grid.6936.a) (ISNI: 0000 0001 2322 2966) 
700 1 |a Fischer, David S.  |u Department of Computational Health, Institute of Computational Biology, Helmholtz Munich, Munich, Germany (ROR: https://ror.org/00cfam450) (GRID: grid.4567.0) (ISNI: 0000 0004 0483 2525); Eric and Wendy Schmidt Center at the Broad Institute, Cambridge, MA, USA (ROR: https://ror.org/05a0ya142) (GRID: grid.66859.34) (ISNI: 0000 0004 0546 1623) 
700 1 |a Theis, Fabian J.  |u Department of Computational Health, Institute of Computational Biology, Helmholtz Munich, Munich, Germany (ROR: https://ror.org/00cfam450) (GRID: grid.4567.0) (ISNI: 0000 0004 0483 2525); TUM School of Computation, Information and Technology, Technical University of Munich, Munich, Germany (ROR: https://ror.org/02kkvpp62) (GRID: grid.6936.a) (ISNI: 0000 0001 2322 2966); TUM School of Life Sciences Weihenstephan, Technical University of Munich, Munich, Germany (ROR: https://ror.org/02kkvpp62) (GRID: grid.6936.a) (ISNI: 0000 0001 2322 2966) 
773 0 |t Nature Machine Intelligence  |g vol. 7, no. 1 (Jan 2025), p. 68-82 
786 0 |d ProQuest  |t Science Database 
856 4 1 |3 Citation/Abstract  |u https://www.proquest.com/docview/3159907619/abstract/embedded/7BTGNMKEMPT1V9Z2?source=fedsrch 
856 4 0 |3 Full Text  |u https://www.proquest.com/docview/3159907619/fulltext/embedded/7BTGNMKEMPT1V9Z2?source=fedsrch 
856 4 0 |3 Full Text - PDF  |u https://www.proquest.com/docview/3159907619/fulltextPDF/embedded/7BTGNMKEMPT1V9Z2?source=fedsrch