A Three-Gene Panel with Prognostic Potential Identified by WGCNA and Machine Learning in Acute Myeloid Leukemia: Evidence from RT-qPCR Validation

Abstract Background Acute myeloid leukemia (AML) represents the most prevalent form of leukemia in adults, arising from the accumulation of multiple genetic mutations and dysregulated gene expression. The substantial molecular heterogeneity characteristic of AML contributes to variability in treatment responses, underscoring the critical importance of understanding the prognostic implications of genetic alterations for guiding therapeutic decisions. Therefore, this study aimed to identify novel genes with prognostic potential in AML patients. Methods Transcriptomic data from adult AML patients were obtained from The Cancer Genome Atlas (TCGA) database. Weighted gene co-expression network analysis (WGCNA) highlighted the turquoise module as the most strongly associated with survival (cor = 0.27, p = 0.002). A total of 851 hub genes from this module were retained for downstream machine learning-based feature selection using Least Absolute Shrinkage and Selection Operator (LASSO), Random Survival Forest (RSF), Elastic Net (ENET), and Adaptive Least Absolute Shrinkage and Selection Operator (AdaLASSO). Based on the mean selection frequency across all four algorithms, the top nine genes, LRCH4, ACOT7, TXN2, GCDH, GIGYF1, INPP5E, MVD, FIBP, and PDLIM7, were retained as the final candidate gene panel. Expression patterns of these genes were validated in an independent AML cohort (GSE37642, GPL96). Results Among them, ACOT7, TXN2, and PDLIM7 showed significant differential expression with concordant direction across cohorts, all exhibiting upregulation (positive log₂FC) in the validation dataset. RT-qPCR validation confirmed significant upregulation of PDLIM7 (p = 0.016) and an increased expression trend of TXN2 (p = 0.710) in AML samples, consistent with bioinformatics predictions, while ACOT7 exhibited significant downregulation (p = 0.038), contrasting with the upregulation observed in the discovery dataset. Conclusions In conclusion, based on significant differences in gene expression and bioinformatics analyses, our study identified three potential prognostic candidate genes that warrant further investigation in prognostic studies.

Authors

Institutions

Publication Details

Journal
Research Square
Published
2026-11-12
DOI
https://doi.org/10.21203/rs.3.rs-11254153/v1
Primary Topic
Acute Myeloid Leukemia Research
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

A Three-Gene Panel with Prognostic Potential Identified by WGCNA and Machine Learning in Acute Myeloid Leukemia: Evidence from RT-qPCR Validation

Leili Tapak, Saeid Afshar, Hassan Rafieemehr, Mohammad Hossein Mohammadi et al.
Research Square
Acute Myeloid Leukemia Research
preprint

A Three-Gene Panel with Prognostic Potential Identified by WGCNA and Machine Learning in Acute Myeloid Leukemia: Evidence from RT-qPCR Validation

Leili Tapak, Saeid Afshar, Hassan Rafieemehr, Mohammad Hossein Mohammadi, Melika Asayesh, Mohammad Reza Safari, Mohammad Rafiee
preprint en

Abstract

Abstract Background Acute myeloid leukemia (AML) represents the most prevalent form of leukemia in adults, arising from the accumulation of multiple genetic mutations and dysregulated gene expression. The substantial molecular heterogeneity characteristic of AML contributes to variability in treatment responses, underscoring the critical importance of understanding the prognostic implications of genetic alterations for guiding therapeutic decisions. Therefore, this study aimed to identify novel genes with prognostic potential in AML patients. Methods Transcriptomic data from adult AML patients were obtained from The Cancer Genome Atlas (TCGA) database. Weighted gene co-expression network analysis (WGCNA) highlighted the turquoise module as the most strongly associated with survival (cor = 0.27, p = 0.002). A total of 851 hub genes from this module were retained for downstream machine learning-based feature selection using Least Absolute Shrinkage and Selection Operator (LASSO), Random Survival Forest (RSF), Elastic Net (ENET), and Adaptive Least Absolute Shrinkage and Selection Operator (AdaLASSO). Based on the mean selection frequency across all four algorithms, the top nine genes, LRCH4, ACOT7, TXN2, GCDH, GIGYF1, INPP5E, MVD, FIBP, and PDLIM7, were retained as the final candidate gene panel. Expression patterns of these genes were validated in an independent AML cohort (GSE37642, GPL96). Results Among them, ACOT7, TXN2, and PDLIM7 showed significant differential expression with concordant direction across cohorts, all exhibiting upregulation (positive log₂FC) in the validation dataset. RT-qPCR validation confirmed significant upregulation of PDLIM7 (p = 0.016) and an increased expression trend of TXN2 (p = 0.710) in AML samples, consistent with bioinformatics predictions, while ACOT7 exhibited significant downregulation (p = 0.038), contrasting with the upregulation observed in the discovery dataset. Conclusions In conclusion, based on significant differences in gene expression and bioinformatics analyses, our study identified three potential prognostic candidate genes that warrant further investigation in prognostic studies.

Research Square
Hamedan University of Medical Sciences (IR), Shahid Beheshti University of Medical Sciences (IR)
Good health and well-being
Acute Myeloid Leukemia Research
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