Machine learning for pan-cancer classification based on RNA sequencing data - Mini Review

Aug 4, 2026

Despite advances in cancer diagnostics, cancers of unknown primary (CUP) remain a major clinical challenge because the tissue of origin (TOO) cannot be identified at presentation, limiting optimal treatment choices. This mini review evaluates 20 recent machine-learning studies that predict TOO from RNA sequencing data, focusing on reproducibility, interpretability, and robustness. It compares reported performance on independent datasets and analyzes identification of key features, dataset issues for training and testing, clinical usefulness, and directions for future improvement.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages7

Reading Time18 min

AccessFree to download

Introduction

CUP and clinical challenge

NGS initiatives and TOO prediction methods

Motivation for machine-learning review

Evaluation scope and themes

TYPE Mini Review PUBLISHED 10 November 2023 DOI 10.3389/fmolb.2023.1285795 OPEN ACCESS EDITED BY Muhammad Tahir Khan, University of Lahore, Pakistan REVIEWED BY Marcelo Cardoso Dos Reis Melo, Auburn University, United States Athar Shaﬁq, Shanghai Jiao Tong University, China *CORRESPONDENCE Rosa Karlić, [rosa@bioinfo.hr](rosa@bioinfo.hr) RECEIVED 30 August 2023 ACCEPTED 30 October 2023 PUBLISHED 10 November 2023 CITATION Štancl P and Karlić R (2023), Machine learning for pan-cancer classiﬁcation based on RNA sequencing data. Front. Mol. Biosci. 10:1285795 . doi: 10.3389/fmolb.2023.1285795 COPYRIGHT © 2023 Štancl and Karlić . This is an openaccess article distributed under the terms of the Creative Commons Attribution License (CC BY) . The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms. Machine learning for pan-cancer classiﬁcation based on RNA sequencing data Paula Štancl and Rosa Karlić * Bioinformatics Group, Division of Molecular Biology, Department of Biology, Faculty of Science, University of Zagreb, Zagreb, Croatia Despite recent improvements in cancer diagnostics, 2%-5% of all malignancies are still cancers of unknown primary (CUP), for which the tissue-of-origin (TOO) cannot be determined at the time of presentation. Since the primary site of cancer leads to the choice of optimal treatment, CUP patients pose a signiﬁcant clinical challenge with limited treatment options. Data produced by large-scale cancer genomics initiatives, which aim to determine the genomic, epigenomic, and transcriptomic characteristics of a large number of individual patients of multiple cancer types, have led to the introduction of various methods that use machine learning to predict the TOO of cancer patients. In this review, we assess the reproducibility, interpretability, and robustness of results obtained by 20 recent studies that utilize different machine learning methods for TOO prediction based on RNA sequencing data, including their reported performance on independent data sets and identiﬁcation of important features. Our review investigates the strengths and weaknesses of different methods, checks the correspondence of their results, and identiﬁes potential issues with datasets used for model training and testing, assessing their potential usefulness ina clinical setting and suggesting future improvements. KEYWORDS cancer of unknown primary, tissue of origin, cancer classiﬁcation, machine learning, RNA sequencing 1 Introduction Cancer is the leading cause of death worldwide, and the overall burden of cancer incidence and mortality is expected to increase due to the growing population, aging inhabitants, and changes in the prevalence of risk factors (Sung et al., 2021) . Key factors in reducing cancer incidence and improving the survival of cancer patients include prevention, early detection, and the availability of appropriate treatment. Despite the recent advances in cancer diagnostics, cancers of unknown primary (CUP), in which the tissue-of-origin (TOO) cannot be identiﬁed at the time of presentation, still constitute approximately 2%–5% of all malignancies. Since the primary site of cancer determines the choice of optimal treatment, CUP patients present a signiﬁcant clinical challenge with limited treatment options. Although a fraction of CUP patients can be assigned the correct TOO and receive appropriate treatment based on the analysis of clinical, imaging, or histopathological data, this is still not the case for the majority of CUP patients, who then face a less favorable prognosis (Binder et al., 2018; Rassy and Pavlidis, 2020) . The development of array-and next-generation sequencing (NGS)

FAQs

Why is tissue-of-origin (TOO) prediction important for CUP patients?

TOO influences selection of optimal therapy. For CUP patients, inability to determine the primary site leads to limited treatment options and poorer prognosis.

What does the review evaluate about machine-learning methods?

The review assesses reproducibility, interpretability, and robustness, including performance on independent datasets and how important features are identified.

Which data type is central to the reviewed methods?

RNA sequencing data is the main input, used to predict the tissue of origin across multiple cancer types.

Aug 4, 2026

Upload by Lucas Martin