Implementation and Evaluation of Support Vector Machine-Based Models for Cancer Detection Using Multi-Omic Data: A Systematic Review

Introduction

Cancer is a major source of mortality and morbidity all over the world that has caused more than 19 million new cases and nearly 10 million deaths in 2020. Although there are so many advances in cancer diagnosis, previous methods such as imaging and serum biomarkers more often lack the necessary sensitivity and specificity, particularly for early-stage detection.

However, most of the studies depend on internal validation that increases concerns about the generalizability of these findings. To improve the dependability of SVM applications in clinical fields, the review emphasizes the necessity of external validation and established techniques.

Due to all the things mixing AI with omics technology suggests a hopeful way to improve cancer detection, that could end up in better results and more affordable medical treatments.

Method

This systematic review was conducted using the PRISMA2020 principles and registered on The Open Science Framework. A comprehensive search of several databases was conducted, including PubMed/MEDLINE, Scopus, Google Scholar, and Web Of Science. Data was screened using RAYYAN.ai, which uses artificial intelligence methods to help with decision-making and screening. All original English-language studies that employed SVM to build a model for diagnosing a type human malignancy were included. The full text of the articles was extracted, and the quality of the articles and risk of bias were assessed using the PROBAST tool.

Result

A total of 104 studies were identified, of which 99 articles have been included after 5 were excluded because full text was unavailable. The studies covered various types of omics, such as proteomics (41 studies), transcriptomics (30 studies), genomics (19 studies), metabolomics (11 studies), epigenomics (4 studies), radiomics (2 studies), immunomics (1 study), and multi-omics (8 studies). 63 studies were internally validated, and 29 were externally validated; however, 2 studies were both internally and externally validated.

Conclusion

The review of 99 studies on Support Vector Machine-based models highlights their potential in improving cancer diagnosis. The study emphasizes the importance of proteomics studies in understanding tumor biology and developing effective diagnostic methods. However, concerns about their generalizability and trustworthiness in medical settings persist.