In molecular biology, a batch effect occurs when non-biological factors in an experiment cause changes in the data produced by the experiment. Such effects can lead to inaccurate conclusions when their causes are correlated with one or more outcomes of interest in an experiment. They are common in many types of high-throughput sequencing experiments, including those using microarrays, mass spectrometers,[1] and single-cell RNA-sequencing data.[2] They are most commonly discussed in the context of genomics and high-throughput sequencing research, but they exist in other fields of science as well.[1]
Definitions
editMultiple definitions of the term "batch effect" have been proposed in the literature. Lazar et al. (2013) noted, "Providing a complete and unambiguous definition of the so-called batch effect is a challenging task, especially because its origins and the way it manifests in the data are not completely known or not recorded." Focusing on microarray experiments, they propose a new definition based on several previous ones: "[T]he batch effect represents the systematic technical differences when samples are processed and measured in different batches and which are unrelated to any biological variation recorded during the MAGE [microarray gene expression] experiment."[3]
Causes
editMany potentially variable factors have been identified as potential causes of batch effects, including the following:
Correction
editVarious statistical techniques have been developed to attempt to correct for batch effects in high-throughput experiments. These techniques are intended for use during the stages of experimental design and data analysis. They have historically mostly focused on genomics experiments, and have only recently begun to expand into other scientific fields such as proteomics.[5] One problem associated with such techniques is that they may unintentionally remove actual biological variation.[6] Some techniques that have been used to detect and/or correct for batch effects include the following:
- For microarray data, linear mixed models have been used, with confounding factors included as random intercepts.[7]
- In 2007, Johnson et al. proposed an empirical Bayesian technique for correcting for batch effects. This approach represented an improvement over previous methods in that it could be effectively used with small batch sizes.[4]
- In 2012, the sva software package was introduced. It includes multiple functions to adjust for batch effects, including the use of surrogate variable estimation, which had previously been shown to improve reproducibility and reduce dependence in high-throughput experiments.
- Haghverdi et al. (2018) proposed a technique designed for single-cell RNA-seq data, based on the detection of mutual nearest neighbors in the data.[2]
- Papiez et al. (2019) proposed a dynamic programming algorithm to identify batch effects of unknown value in high-throughput data.[8]
- Voß et al. (2022) proposed an algorithm called HarmonizR which enables data harmonization across independent proteomic datasets with appropriate handling of missing values.[9]
References
edit- ^ a b c d e Leek, Jeffrey T.; Scharpf, Robert B.; Bravo, Héctor Corrada; Simcha, David; Langmead, Benjamin; Johnson, W. Evan; Geman, Donald; Baggerly, Keith; Irizarry, Rafael A. (October 2010). "Tackling the widespread and critical impact of batch effects in high-throughput data". Nature Reviews Genetics. 11 (10): 733–739. doi:10.1038/nrg2825. ISSN 1471-0056. PMC 3880143. PMID 20838408.
- ^ a b Haghverdi, Laleh; Lun, Aaron T L; Morgan, Michael D; Marioni, John C (May 2018). "Batch effects in single-cell RNA-sequencing data are corrected by matching mutual nearest neighbors". Nature Biotechnology. 36 (5): 421–427. doi:10.1038/nbt.4091. ISSN 1087-0156. PMC 6152897. PMID 29608177.
- ^ Leek, Jeffrey T.; Johnson, W. Evan; Parker, Hilary S.; Jaffe, Andrew E.; Storey, John D. (2012-03-15). "The sva package for removing batch effects and other unwanted variation in high-throughput experiments". Bioinformatics. 28 (6): 882–883. doi:10.1093/bioinformatics/bts034. ISSN 1460-2059. PMC 3307112. PMID 22257669.
- ^ a b c d Johnson, W. Evan; Li, Cheng; Rabinovic, Ariel (2007-01-01). "Adjusting batch effects in microarray expression data using empirical Bayes methods". Biostatistics. 8 (1): 118–127. doi:10.1093/biostatistics/kxj037. ISSN 1468-4357. PMID 16632515.
- ^ Čuklina, Jelena; Pedrioli, Patrick G. A.; Aebersold, Ruedi (2020). Review of Batch Effects Prevention, Diagnostics, and Correction Approaches. Methods in Molecular Biology. Vol. 2051. pp. 373–387. doi:10.1007/978-1-4939-9744-2_16. ISBN 978-1-4939-9743-5. ISSN 1940-6029. PMID 31552638. S2CID 202760910.
- ^ Goh, Wilson Wen Bin; Wang, Wei; Wong, Limsoon (June 2017). "Why Batch Effects Matter in Omics Data, and How to Avoid Them". Trends in Biotechnology. 35 (6): 498–507. doi:10.1016/j.tibtech.2017.02.012. PMID 28351613.
- ^ Espín-Pérez, Almudena; Portier, Chris; Chadeau-Hyam, Marc; van Veldhoven, Karin; Kleinjans, Jos C. S.; de Kok, Theo M. C. M. (2018-08-30). Krishnan, Viswanathan V. (ed.). "Comparison of statistical methods and the use of quality control samples for batch effect correction in human transcriptome data". PLOS ONE. 13 (8): e0202947. Bibcode:2018PLoSO..1302947E. doi:10.1371/journal.pone.0202947. ISSN 1932-6203. PMC 6117018. PMID 30161168.
- ^ Papiez, Anna; Marczyk, Michal; Polanska, Joanna; Polanski, Andrzej (2019-06-01). Berger, Bonnie (ed.). "BatchI: Batch effect Identification in high-throughput screening data using a dynamic programming algorithm". Bioinformatics. 35 (11): 1885–1892. doi:10.1093/bioinformatics/bty900. ISSN 1367-4803. PMC 6546123. PMID 30357412.
- ^ Voß, Hannah; Schlumbohm, Simon; Barwikowski, Philip; Wurlitzer, Marcus; Dottermusch, Matthias; Neumann, Philipp; Schlüter, Hartmut; Neumann, Julia E.; Krisp, Christoph (2022-06-20). "HarmonizR enables data harmonization across independent proteomic datasets with appropriate handling of missing values". Nature Communications. 13 (1): 3523. doi:10.1038/s41467-022-31007-x. ISSN 2041-1723. PMC 9209422. PMID 35725563.