EUVIMEDEUVIMED entdecken
Open Access · CC BY-NC-SA 4.0

Biologically Coherent MicroRNA Signature for Machine-Learning-Based Colorectal Cancer Detection

Rebaz M. Mustafa, Kamaran Abdoulrahman, Mohammad Hamid

ARO-THE SCIENTIFIC JOURNAL OF KOYA UNIVERSITY · 08.11.2026 · 14 (2) · 127-139

DOI: 10.14500/aro.13036 ↗

Kurz gefasst

Fachartikel in der Zeitschrift ARO-THE SCIENTIFIC JOURNAL OF KOYA UNIVERSITY, online erschienen am 08.11.2026. Verfasst von Rebaz M. Mustafa und 2 weiteren Autor:innen. Die Arbeit nennt 32 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-SA 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Colorectal cancer (CRC) biomarker discovery is frequently limited by signature instability, methodological heterogeneity, and insufficient external validation. This study aimed to develop a reproducible bioinformatics and machine learning framework for identifying a compact mature-microRNA (miRNA) signature capable of discriminating CRC tissues from matched normal controls. Small RNA-seq data from a paired discovery cohort (PRJNA387172; 104 tumor–normal pairs) and three independent external high-throughput sequencing cohorts (GSE125961, GSE183437, and GSE236991) were processed through uniform pre-processing, mature-miRNA quantification, counts per-million filtering, and TMM/logCPM normalization. Within the discovery cohort, pair-preserving 80:20 partitioning was combined with repeated stratified five-fold cross-validation with 10 repeats, train-only differential-expression analysis, and consensus feature selection using Elastic Net, Boruta, and random forest recursive feature elimination. This framework identified a stable locked three-miRNA signature comprising hsa-miR-135b-5p, hsa-miR-378a-5p, and hsa-miR-549a-5p. A random forest classifier trained on this panel achieved perfect discrimination in the discovery partitions, with an area under the curve (AUC) of 1.000 in both the training and independent test subsets, and showed preliminary external performance, with panel AUCs of 0.917, 1.000, and 0.840 across three small validation cohorts and 0.961 in pooled external analysis. Expression direction was preserved across datasets, with miR-135b-5p and miR-549a-5p upregulated and miR-378a-5p downregulated in tumors. Downstream target integration analysis recognized 401 overlap-supported miRNA-target interactions spanning 393 unique genes, with enrichment in Wnt, Hippo, transforming growth factor beta, focal adhesion, and CRC-related pathways. These findings support a biologically coherent three-miRNA random forest model for CRC biomarker discovery and prioritize candidates for future independent and experimental validation.

Zitieren

APA

Mustafa, R. M., Abdoulrahman, K., & Hamid, M. (2026). Biologically Coherent MicroRNA Signature for Machine-Learning-Based Colorectal Cancer Detection. ARO-THE SCIENTIFIC JOURNAL OF KOYA UNIVERSITY, 14(2), 127-139. https://doi.org/10.14500/aro.13036

Vancouver

Mustafa RM, Abdoulrahman K, Hamid M. Biologically Coherent MicroRNA Signature for Machine-Learning-Based Colorectal Cancer Detection. ARO-THE SCIENTIFIC JOURNAL OF KOYA UNIVERSITY. 2026;14(2):127-139. doi:10.14500/aro.13036

BibTeX
@article{Mustafa2026,
  author = {Mustafa, Rebaz M. and Abdoulrahman, Kamaran and Hamid, Mohammad},
  title = {Biologically Coherent MicroRNA Signature for Machine-Learning-Based Colorectal Cancer Detection},
  journal = {ARO-THE SCIENTIFIC JOURNAL OF KOYA UNIVERSITY},
  year = {2026},
  volume = {14},
  number = {2},
  pages = {127--139},
  doi = {10.14500/aro.13036}
}
Kostenlos starten →