Dealing with dimensionality: the application of machine learning to multi-omics data

Abstract

Motivation Machine learning (ML) methods are motivated by the need to automate information extraction from large data sets in order to support human users in data-driven tasks. This is an attractive approach for integrative joint analysis of vast amounts of omics data produced in next generation sequencing and other -omics assays. A systematic assessment of the current literature can help to identify key trends and potential gaps in methodology and applications. We surveyed the literature on ML multi-omic data integration and quantitatively explored the goals, techniques and data involved in this field. We were particularly interested in examining how researchers use ML to deal with the volume and complexity of these datasets. Results Our main finding is that the methods used are those that address the challenges of datasets with few samples and many features. Dimensionality reduction methods are used to reduce the feature count alongside models that can also appropriately handle relatively few samples. Popular techniques include autoencoders, random forests and support vector machines. We also found that the field is heavily influenced by the use of The Cancer Genome Atlas data set, which is accessible and contains many diverse experiments. Availability All data and processing scripts are available at this GitLab repository: https://gitlab.com/polavieja_lab/ml_multi-omics_review/ or in Zenodo: https://doi.org/10.5281/zenodo.7361807 Supplementary information Supplementary data are available at Bioinformatics online.

Genetic risk assessment based on association andprediction studies

“…More so, using the entire feature space without considering the relevance of the individual features hinders achieving an optimal model performance. Given that genomic datasets suffer from the curse of dimensionality 6,8,[29][30][31][32] , it is crucial to eliminate irrelevant features and retain only the most informative variants related to the phenotype under investigation. Removing noise from the data improves models' accuracy and reliability, thereby gaining a deeper understanding of the genetic mechanisms underlying risk susceptibility.…”

“…Machine learning (ML) is a widely accepted methodical framework in analyzing high-dimensional and complex data 6,8,[29][30][31][32] , owing to its unparalleled ability to handle high-volume data and uncover implicit and nonlinear patterns that are pertinent for predictive modeling. By selecting a minimum subset of individually relevant and neighboring features while minimizing information loss 33 , ML captures complex interactions, leading to the identification of highly-predictive features.…”

Microbiome-based risk prediction in incident heart failure: a community challenge

“…A large number of features is a common problem in predictive machine learning models 38 . Thus, similar to a previous study of microbiome-based prediction of all-cause mortality 21 , we found the approach to compress the data into a smaller number of informative characteristics by aggregating species into co-abundance networks provided a strategy for feature engineering in microbiome-based risk prediction.…”

Survival Prediction Landscape: An In-Depth Systematic Literature Review on Activities, Methods, Tools, Diseases, and Databases

“…These articles primarily aim to summarize the latest trends and developments in data modalities, feature engineering methods, and AI models specifically related to survival prediction. However, the focus of these reviews is often constrained to either a singular disease or multiple subtypes of cancer, highlighting a limited scope within the broader landscape of survival prediction research 37,[44][45][46][47][48] . More comprehensive details about the scope of existing review articles in terms of contributions and drawbacks are summarised in Table 1 and section .…”