Integrative analysis of epigenetic subtypes in acute myeloid Leukemia: A multi-center study combining machine learning for prognostic and therapeutic insights.
<h4>Background</h4>Acute Myeloid Leukemia (AML) exhibits significant heterogeneity in clinical outcomes, yet current prognostic stratification systems based on genetic alterations alone cannot fully capture this complexity. This study aimed to develop an integrated epigenetic-based class...
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Public Library of Science (PLoS)
2025-01-01
|
| Series: | PLoS ONE |
| Online Access: | https://doi.org/10.1371/journal.pone.0324380 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1850160089438093312 |
|---|---|
| author | Jincan Li Shengyue Wang |
| author_facet | Jincan Li Shengyue Wang |
| author_sort | Jincan Li |
| collection | DOAJ |
| description | <h4>Background</h4>Acute Myeloid Leukemia (AML) exhibits significant heterogeneity in clinical outcomes, yet current prognostic stratification systems based on genetic alterations alone cannot fully capture this complexity. This study aimed to develop an integrated epigenetic-based classification system and evaluate its prognostic value.<h4>Methods</h4>We performed multi-omics analysis on five independent cohorts totaling 1,103 AML patients. The Cancer Genome Atlas-Acute Myeloid Leukemia (TCGA-LAML) cohort (n = 83) provided comprehensive multi-omics data including DNA methylation profiles (Illumina 450K platform), RNA sequencing (mRNA, lncRNA, and miRNA), and somatic mutation profiles. The BEAT (n = 649), TARGET (n = 156), GSE12417 (n = 79), and GSE37642 (n = 136) cohorts contributed transcriptome data. Molecular subtypes were identified using empirical Bayes-based clustering on the TCGA cohort. LSC17 scores were calculated using a validated 17-gene expression signature. A random survival forest model was developed integrating molecular features with LSC17 scores, validated across all cohorts. Immune microenvironment analysis employed multiple deconvolution methods (ESTIMATE, CIBERSORT, xCell) and pathway analysis (GSVA, GSEA). Drug sensitivity was predicted using the pRRophetic algorithm with GDSC database reference.<h4>Results</h4>Multi-omics integration revealed two molecularly distinct AML subtypes with significant survival differences (CS2 vs CS1, P < 0.001). The random survival forest model, incorporating 20 key epigenetic features (including CPNE8, CD109, and CHRDL1) and LSC17 scores, achieved superior prognostic accuracy (C-index: 0.72-0.78) across validation cohorts. Both epigenetic risk score (HR = 2.45, 95%CI: 1.86-3.24) and LSC17 score (HR = 1.89, 95%CI: 1.42-2.51) maintained independent prognostic value in multivariate analysis. Integration of both scores in a nomogram improved 1-, 3-, and 5-year survival predictions (C-index: 0.81). High-risk patients exhibited distinct immune profiles with elevated M2 macrophages (1.8-fold) and Tregs (2.3-fold), while low-risk patients showed enhanced NK cell activity (2.1-fold). Drug sensitivity analysis identified differential responses to epigenetic regulators (LAQ824, P = 0.000139; MS-275, P = 0.00104) and proteasome inhibitors (Bortezomib, P = 0.00747; MG-132, P = 0.0106) between risk groups.<h4>Conclusions</h4>This integrated classification system combining epigenetic features and stem cell signatures provides new insights into AML heterogeneity and therapeutic targeting. The complementary nature of epigenetic and stem cell-related prognostic factors suggests potential for improved risk stratification in clinical practice. Future prospective validation studies are warranted to confirm these findings. |
| format | Article |
| id | doaj-art-6ba50aaff57f4963b4e294af13481808 |
| institution | OA Journals |
| issn | 1932-6203 |
| language | English |
| publishDate | 2025-01-01 |
| publisher | Public Library of Science (PLoS) |
| record_format | Article |
| series | PLoS ONE |
| spelling | doaj-art-6ba50aaff57f4963b4e294af134818082025-08-20T02:23:15ZengPublic Library of Science (PLoS)PLoS ONE1932-62032025-01-01205e032438010.1371/journal.pone.0324380Integrative analysis of epigenetic subtypes in acute myeloid Leukemia: A multi-center study combining machine learning for prognostic and therapeutic insights.Jincan LiShengyue Wang<h4>Background</h4>Acute Myeloid Leukemia (AML) exhibits significant heterogeneity in clinical outcomes, yet current prognostic stratification systems based on genetic alterations alone cannot fully capture this complexity. This study aimed to develop an integrated epigenetic-based classification system and evaluate its prognostic value.<h4>Methods</h4>We performed multi-omics analysis on five independent cohorts totaling 1,103 AML patients. The Cancer Genome Atlas-Acute Myeloid Leukemia (TCGA-LAML) cohort (n = 83) provided comprehensive multi-omics data including DNA methylation profiles (Illumina 450K platform), RNA sequencing (mRNA, lncRNA, and miRNA), and somatic mutation profiles. The BEAT (n = 649), TARGET (n = 156), GSE12417 (n = 79), and GSE37642 (n = 136) cohorts contributed transcriptome data. Molecular subtypes were identified using empirical Bayes-based clustering on the TCGA cohort. LSC17 scores were calculated using a validated 17-gene expression signature. A random survival forest model was developed integrating molecular features with LSC17 scores, validated across all cohorts. Immune microenvironment analysis employed multiple deconvolution methods (ESTIMATE, CIBERSORT, xCell) and pathway analysis (GSVA, GSEA). Drug sensitivity was predicted using the pRRophetic algorithm with GDSC database reference.<h4>Results</h4>Multi-omics integration revealed two molecularly distinct AML subtypes with significant survival differences (CS2 vs CS1, P < 0.001). The random survival forest model, incorporating 20 key epigenetic features (including CPNE8, CD109, and CHRDL1) and LSC17 scores, achieved superior prognostic accuracy (C-index: 0.72-0.78) across validation cohorts. Both epigenetic risk score (HR = 2.45, 95%CI: 1.86-3.24) and LSC17 score (HR = 1.89, 95%CI: 1.42-2.51) maintained independent prognostic value in multivariate analysis. Integration of both scores in a nomogram improved 1-, 3-, and 5-year survival predictions (C-index: 0.81). High-risk patients exhibited distinct immune profiles with elevated M2 macrophages (1.8-fold) and Tregs (2.3-fold), while low-risk patients showed enhanced NK cell activity (2.1-fold). Drug sensitivity analysis identified differential responses to epigenetic regulators (LAQ824, P = 0.000139; MS-275, P = 0.00104) and proteasome inhibitors (Bortezomib, P = 0.00747; MG-132, P = 0.0106) between risk groups.<h4>Conclusions</h4>This integrated classification system combining epigenetic features and stem cell signatures provides new insights into AML heterogeneity and therapeutic targeting. The complementary nature of epigenetic and stem cell-related prognostic factors suggests potential for improved risk stratification in clinical practice. Future prospective validation studies are warranted to confirm these findings.https://doi.org/10.1371/journal.pone.0324380 |
| spellingShingle | Jincan Li Shengyue Wang Integrative analysis of epigenetic subtypes in acute myeloid Leukemia: A multi-center study combining machine learning for prognostic and therapeutic insights. PLoS ONE |
| title | Integrative analysis of epigenetic subtypes in acute myeloid Leukemia: A multi-center study combining machine learning for prognostic and therapeutic insights. |
| title_full | Integrative analysis of epigenetic subtypes in acute myeloid Leukemia: A multi-center study combining machine learning for prognostic and therapeutic insights. |
| title_fullStr | Integrative analysis of epigenetic subtypes in acute myeloid Leukemia: A multi-center study combining machine learning for prognostic and therapeutic insights. |
| title_full_unstemmed | Integrative analysis of epigenetic subtypes in acute myeloid Leukemia: A multi-center study combining machine learning for prognostic and therapeutic insights. |
| title_short | Integrative analysis of epigenetic subtypes in acute myeloid Leukemia: A multi-center study combining machine learning for prognostic and therapeutic insights. |
| title_sort | integrative analysis of epigenetic subtypes in acute myeloid leukemia a multi center study combining machine learning for prognostic and therapeutic insights |
| url | https://doi.org/10.1371/journal.pone.0324380 |
| work_keys_str_mv | AT jincanli integrativeanalysisofepigeneticsubtypesinacutemyeloidleukemiaamulticenterstudycombiningmachinelearningforprognosticandtherapeuticinsights AT shengyuewang integrativeanalysisofepigeneticsubtypesinacutemyeloidleukemiaamulticenterstudycombiningmachinelearningforprognosticandtherapeuticinsights |