Machine Learning Classification Models using RNA-seq Gene Expression Data for Early- and Late-Stage Kidney Renal Clear Cell Carcinoma

DOI

10.2991/978-94-6463-684-0_17How to use a DOI?

Keywords

RNA-seq; early- and late-stage diseases; Kidney Renal Clear Cell Carcinoma; machine learning; classification models

Abstract

Kidney Renal Clear Cell Carcinoma (KIRC) is one of the most common cancers in the world. With limited ideal testing and screening methods, this disease has been prone to misclassification and poor prognosis leading to late-stage detection and metastasis. One of the few methods that have been developed to address this is the use of biomarkers through gene expression. This study used KIRC TCGA RNA-seq gene expression data to classify early-stage or late-stage KIRC. The expressed genes are the features filtered through Information Gain Feature selection. The identified features were then used to train six (6) Machine Learning (ML) classification models and were evaluated by five (5) performance measures. These models were the Naive Bayes (NB), Sequential Minimal Optimization (SMO), Stochastic Gradient Descent (SGD), Logistic Regression (LR), J48 Classification (J48), and Random Forest (RF) classifiers. The performance measures used were accuracy, sensitivity, specificity, F-value, and the auROC. The features used in training the ML classification models were also subject to biological functional analysis through Gene Ontology (GO) Cellular Component enrichment analysis, GO molecular function enrichment analysis, and the KEGG Pathway Enrichment Analysis. The Machine learning classifier performance evaluation revealed that RF was the best-performing model in early stage KIRC, with a good performance in accuracy and specificity, followed by SGD. In contrast, all models had poor performances with the Late stage KIRC, equating to an overall average performance. The GO biological functional analyses and KEGG pathway analysis also showed the dominance of cell cycle-associated genes. Possibilities of improving future ML Classification model development through modifications in feature selection were idealized from the KEGG results. Its results were able to identify the likelihood of having feature-associated noise due highly expressed metabolic genes by diverse cell types, leading to the poor late-stage KIRC performances of the tested models.

Copyright

© 2025 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - Lawrence S. Macalalad
AU  - Geoffrey A. Solano
AU  - Joey Mark S. Diaz
PY  - 2025
DA  - 2025/04/30
TI  - Machine Learning Classification Models using RNA-seq Gene Expression Data for Early- and Late-Stage Kidney Renal Clear Cell Carcinoma
BT  - Proceedings of the  Workshop on Computation: Theory and Practice (WCTP 2024)
PB  - Atlantis Press
SP  - 264
EP  - 280
SN  - 2589-4900
UR  - https://doi.org/10.2991/978-94-6463-684-0_17
DO  - 10.2991/978-94-6463-684-0_17
ID  - Macalalad2025
ER  -