Abstract
With the recent advances in molecular biology techniques such as next-generation sequencing, mass-spectrometry, etc., a large omic data is produced. Using such data, the expression levels of thousands of molecular features (genes, proteins, metabolites, etc.) can be quantified and associated with diseases. The fact that multiple omics data contains different types of data and the number of analyzed variables increases the complexity of the models created with machine learning methods. In addition, due to many variables, the investigation of molecular variables associated with diseases is very costly. Therefore, selecting the informative and disease-related molecular features is applicable before model training and evaluation. This feature selection step is essential for obtaining accurate and generalizable models in minimum time with minimum cost. Some current methods used for feature selection are as follows: recursive feature elimination, information gain, minimum redundancy maximum relevance (mRMR), boruta, altmann, and lasso.
TopIntroduction
Today, there is an increase in data in many areas. With this increase, the number and variety of the variables to be evaluated also increases. The increase in data and variables became a situation that needed to be solved among world problems. In addition, although there is a perception that having too much data in the scientific field, having too much information, correct information, or sufficient information may not be possible. However, it should not be forgotten that there is valuable information in a relatively large amount of data. It should be clear that it can be beneficial to have much data to extract this helpful information. However, performing data analyses to obtain and process this information can be difficult. In addition, its existence is a problem called the curse of data dimensionality (Verkeysen M. and François D., 2005). High-dimensional data sets, where these problems are most common, are used successfully in multiple fields such as genetics, pharmacology, toxicology, nutrition, and genetics. The use of these high-dimensional data allows one to examine biology systems, cellular metabolism, and disease etiologies in more detail. However, the number of samples (n) of these data is considerably lower than the number of variables (p) and the heterogeneity of the data, the missing observations in the data as a result of the use of high-output technology, limits the use of traditional methods that can be used in this field. Therefore, there is a need for the clinical understanding of the biological system based on research and machine learning, and statistical learning methods to analyze this clinical information statistically (Hastie et al., 2009). Several studies are show that machine learning methods are used and applied successfully in studies carried out in this field. Some of these studies are listed in Table 1.
Table 1. Some studies using feature selection
| Datasets | | Methods | References |
| Ovarian cancer | Classification | MKL | Wilson et al. | 2019 | |
| Breast cancer | Classification | MKL | Tao et al. | 2019 | |
| Gene expression | Classification | SVM | Golub et al. | 1999 | |
| Gut microbiota | Classification | RF | Franzosa et al. | 2019 | |
| Colon cancer | Classification | SVM | Moler et al. | 2000 | |
| Ovarian, leukemia, colon | Classification | SVM | Furey et al. | 2000 | |
MKL: Multiple Kernel Learning, SVM: Support Vector Machine, RF:Random Forest
Key Terms in this Chapter
Next Generation Sequencing: It is a new technology for DNA, RNA sequencing, and variant/mutation detection. Next-generation sequencing is also called large parallel sequencing or second-generation sequencing. Next-generation sequencing is a method of simultaneously sequencing DNA fragments (or complementary DNA).
Filter Methods: Filter methods use statistical techniques to evaluate the relationship between input and target variables. These scores are used to choose (filter) those input variables that will be used in the model.
Systems Biology: It is the computational and mathematical analysis and modeling of complex biological systems. It is an approach in biomedical research to understand the larger picture, be it at the level of the organism, tissue, or cell, by putting its pieces together.
Mass Spectrometry: It is an analytical technique used to measure the mass-to-charge ratio of ions.
Multi-Omics: It is a new approach where the data sets of different omic (metabolomic, genomic, proteomic, transcriptomic, etc.) groups are combined during analysis.
Metabolomics: It detects, quantifies, and identifies small molecule metabolites emerging from lipids, carbohydrates, vitamins, hormones, and other cell components in tissues, cells, and physiological fluids in a certain time high-throughput technologies.
Gene Expression: It is the process by which information from a gene is used to synthesize a functional gene product that enables it to produce end products, protein or non-coding RNA, and ultimately affect a phenotype as the final effect.
Feature Selection: It is defined as selecting the best subset that can represent the original dataset.