Generalizable and Interpretable Ensemble Framework for Classification of Bacterial and Viral Infections Using 41-Gene Host RNA Signature
8th International Congress on Human-Computer Interaction, Optimization and Robotic Applications, ICHORA 2026, Ankara, Türkiye, 21 - 23 Mayıs 2026, (Tam Metin Bildiri)
- Yayın Türü: Bildiri / Tam Metin Bildiri
- Doi Numarası: 10.1109/ichora69329.2026.11537061
- Basıldığı Şehir: Ankara
- Basıldığı Ülke: Türkiye
- Anahtar Kelimeler: artificial neural network, infectious diseases, machine learning, soft voting ensemble, transcriptomic data
- Dokuz Eylül Üniversitesi Adresli: Evet
Özet
The accurate etiological diagnosis of infectious diseases using high-dimensional transcriptomic data is often prevented by biological noise and cross-platform heterogeneity. To overcome this limitation, we present a robust, platformindependent machine learning approach developed from a large multicohort data set containing $\mathbf{1, 8 9 8}$ samples from $\mathbf{1 1}$ microarray platforms. Following referenceComBat normalization to eliminate technical batch effects, we employed the One-vs-Rest (OvR) ensemble feature selection strategy. This approach successfully transformed thousands of transcripts into a highly discriminative, 41-gene multi-class signature. For the classification, we built a soft voting ensemble that combines Support Vector Machines (SVM), LightGBM, and Artificial Neural Network (ANN) algorithms, with each hyperparameter dynamically optimized using the Optuna framework. During independent external validation, the recommended ensemble model performed well, achieving a multiclass AUC value of 0.913 in the GSE72809 cohort. Furthermore, in a viral-healthy test using the GSE77087 dataset-which included only viral and healthy individuals-the model achieved an AUC value of 0.938, which can be considered high. These findings demonstrate that our optimized 41-gene signature and ensemble model approach provide a reliable, scalable, and robust foundation for the accurate and rapid etiological diagnosis of infectious diseases in clinical conditions.