About this Research Topic
Biomarker signatures derived from big data in medicine possess the potential to provide more sensitive, specific, and robust predictions of disease-associated outcomes, compared to classic single-molecular biomarkers. When dealing with big data, machine learning is a commonly used and powerful statistical method. Nevertheless, only a few conventional machine learning methods, such as LASSO and logistic regression models have gained widespread utilization in relevant bioinformatics analyses of biomarker discovery. In this research topic, we call for novel machine learning methods, particularly feature selection algorithms that lead to more effective biomarker signatures with superior accuracy and better biological relevance, and their applications to identify diagnostic and/or prognostic biomarker signatures for complex diseases, which would pave a way towards precision medicine.
This Research Topic aims to address the limited utilization of advanced machine learning methods in the area of biomarker discovery. Its focus is on boosting the development and application of relevant innovative machine learning methods to analyze big data in medicine, including omics data, laboratory data, medical image data, and others. Special emphasis is placed on feature selection algorithms that incorporate biological pathway information as a priori to facilitate the process of selecting relevant features/markers. We believe that developing novel machine learning methods or seamlessly adapting existing algorithms for biomarker discovery in complex diseases can aid the successful development of clinically valuable diagnostic and prognostic tests, leading to more “precision” clinical practice and regimes.
The research topic encompasses a variety of themes, including but not limited to:
1) Development and application of machine learning methods (e.g., feature selection algorithms) to identify diagnostic and prognostic signatures, particularly for single-cell RNA-Seq data, spatial transcriptomics data, and medical image data;
2) Development or adaptation of advanced methods that incorporate feature-to-feature interaction information prior to aid in the feature selection process, resulting in biomarkers with better biological interpretation and more satisfactory performance;
3) Development of interpretable deep learning methods that equip deep neural networks with feature selection or causal inference strategies to turn a “black box” into a “white box”;
4) Integrative analyses of multiple-view or multiple-modal or multiple omics data, particularly those based on the pathway\network analysis methods to construct the biomarker signatures for complex diseases;
5) Development of statistical metrics that can provide a better evaluation of the performance of identified signatures.
This Research Topic aims to address the limited utilization of advanced machine learning methods in the area of biomarker discovery. Its focus is on boosting the development and application of relevant innovative machine learning methods to analyze big data in medicine, including omics data, laboratory data, medical image data, and others. Special emphasis is placed on feature selection algorithms that incorporate biological pathway information as a priori to facilitate the process of selecting relevant features/markers. We believe that developing novel machine learning methods or seamlessly adapting existing algorithms for biomarker discovery in complex diseases can aid the successful development of clinically valuable diagnostic and prognostic tests, leading to more “precision” clinical practice and regimes.
The research topic encompasses a variety of themes, including but not limited to:
1) Development and application of machine learning methods (e.g., feature selection algorithms) to identify diagnostic and prognostic signatures, particularly for single-cell RNA-Seq data, spatial transcriptomics data, and medical image data;
2) Development or adaptation of advanced methods that incorporate feature-to-feature interaction information prior to aid in the feature selection process, resulting in biomarkers with better biological interpretation and more satisfactory performance;
3) Development of interpretable deep learning methods that equip deep neural networks with feature selection or causal inference strategies to turn a “black box” into a “white box”;
4) Integrative analyses of multiple-view or multiple-modal or multiple omics data, particularly those based on the pathway\network analysis methods to construct the biomarker signatures for complex diseases;
5) Development of statistical metrics that can provide a better evaluation of the performance of identified signatures.
Keywords: Feature selection, High-dimensional data, Diagnostic signature, Prognosis, Complex disease
Important Note: All contributions to this Research Topic must be within the scope of the section and journal to which they are submitted, as defined in their mission statements. Frontiers reserves the right to guide an out-of-scope manuscript to a more suitable section or journal at any stage of peer review.
