Genes That Reliably Differentiate Liver‐Fibrosis Stages Identified Using a Novel Integrated Statistical and Machine‐Learning Algorithm

Accurate prediction of stage of liver fibrosis is essential for risk stratification and management of metabolic dysfunction-associated steatotic liver disease (MASLD). However, previously reported genomic and transcriptomic predictors of fibrosis-stage have shown limited reproducibility. We aimed to identify robust biomarkers predictive of the stage (mild [F1] or advanced [F3 or F4]) of fibrosis. Transcriptomic data were curated from public repositories using stringent quality-control procedures to generate high-quality whole-transcriptome profiles from 278 patients (F1: n = 139; F3: n = 109; F4: n = 30). Samples were randomly divided into equal-sized discovery and validation cohorts. To distinguish mild from advanced fibrosis, we initially applied linear discriminant analysis (LDA) and support vector machine (SVM) classification with recursive backward feature elimination (RFE). These approaches yielded suboptimal results; both LDA and SVM-RFE required large gene sets-74 and 88 genes, respectively-but with limited percentages of correct classification of patients with mild and advanced fibrosis-69.78% and 79.86%, respectively. We, therefore, developed a novel integrative algorithm combining LDA pre-screening and SVM-Elastic Net with cross validation. This algorithm identified 20 genes whose expression-levels-that were not significantly correlated-achieved an accuracy of 85.61% (AUC = 0.8892) in classifying patients with mild and advanced stages of liver fibrosis in an independent dataset. We also assessed classification performance using a Multilayer Perceptron and an SVM with an RBF kernel-both non-linear classifiers-that achieved accuracies of 82.3% (average) and 81.29%, respectively; both accuracies are less than that achieved by SVM-EN. Of the 20 genes identified by us, previous studies have strongly implicated 10 genes in liver fibrosis, while the remaining 10 genes are novel in the context of involvement in liver fibrosis and require confirmation.

Authors

Institutions

Publication Details

Journal
Genetic Epidemiology
Published
2026-11-16
DOI
https://doi.org/10.1002/gepi.70064
Primary Topic
Liver Disease Diagnosis and Treatment
Type
article
Field-Weighted Citation Impact
0.00

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Genes That Reliably Differentiate Liver‐Fibrosis Stages Identified Using a Novel Integrated Statistical and Machine‐Learning Algorithm

A Chatterjee, Nikhil Ranjan Pal, Partha P. Majumder, Vivekananda Mallik et al.
5 links
Genetic Epidemiology
Liver Disease Diagnosis and Treatment
article

Genes That Reliably Differentiate Liver‐Fibrosis Stages Identified Using a Novel Integrated Statistical and Machine‐Learning Algorithm

A Chatterjee, Nikhil Ranjan Pal, Partha P. Majumder, Vivekananda Mallik, Pranoy Naskar, Abhijit Chowdhury
article en

Abstract

Accurate prediction of stage of liver fibrosis is essential for risk stratification and management of metabolic dysfunction-associated steatotic liver disease (MASLD). However, previously reported genomic and transcriptomic predictors of fibrosis-stage have shown limited reproducibility. We aimed to identify robust biomarkers predictive of the stage (mild [F1] or advanced [F3 or F4]) of fibrosis. Transcriptomic data were curated from public repositories using stringent quality-control procedures to generate high-quality whole-transcriptome profiles from 278 patients (F1: n = 139; F3: n = 109; F4: n = 30). Samples were randomly divided into equal-sized discovery and validation cohorts. To distinguish mild from advanced fibrosis, we initially applied linear discriminant analysis (LDA) and support vector machine (SVM) classification with recursive backward feature elimination (RFE). These approaches yielded suboptimal results; both LDA and SVM-RFE required large gene sets-74 and 88 genes, respectively-but with limited percentages of correct classification of patients with mild and advanced fibrosis-69.78% and 79.86%, respectively. We, therefore, developed a novel integrative algorithm combining LDA pre-screening and SVM-Elastic Net with cross validation. This algorithm identified 20 genes whose expression-levels-that were not significantly correlated-achieved an accuracy of 85.61% (AUC = 0.8892) in classifying patients with mild and advanced stages of liver fibrosis in an independent dataset. We also assessed classification performance using a Multilayer Perceptron and an SVM with an RBF kernel-both non-linear classifiers-that achieved accuracies of 82.3% (average) and 81.29%, respectively; both accuracies are less than that achieved by SVM-EN. Of the 20 genes identified by us, previous studies have strongly implicated 10 genes in liver fibrosis, while the remaining 10 genes are novel in the context of involvement in liver fibrosis and require confirmation.

Genetic EpidemiologyVol. 50(8)
Indian Statistical Institute (IN)
Indian Council of Medical Research
Good health and well-being
Liver Disease Diagnosis and Treatment
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon