This bioinformatics study combines two-stage feature selection with a neural classifier to identify candidate Mpox-associated genes from two microarray datasets and one RNA-seq dataset, followed by network analyses and molecular docking.
Key findings
- Thirty-three genes overlapped between the DEG and final selected sets. PPI analysis identified ten key genes and regulatory networks highlighted five hubs. Docking suggested interactions involving ATG3, TRIM14 and DUOX1, but these remain computational predictions.
Why this matters globally
The pipeline may accelerate hypothesis prioritisation for Mpox pathogenesis, detection and drug research, provided that findings replicate in independent cohorts and receive laboratory validation.
Thai researcher contribution
Watshara Shoombuatong of Mahidol University contributed to the study, reflecting Thai expertise in machine learning and bioinformatics within an international infectious-disease collaboration.
Limitations to consider
Sample sizes and splitting details require inspection in the full paper. Platform and batch effects may influence results, while feature selection on limited datasets risks overfitting and analytic leakage. Docking does not prove binding, drug efficacy or clinical utility, and no independent clinical or wet-lab validation is reported.