Machine learning and molecular validation reveal the PI3K/AKT/mTOR inflammatory axis for therapeutic targeting in severe COVID-19

Dysregulation of the PI3K/AKT/mTOR signaling axis contributes to COVID-19 severity by amplifying hyperinflammatory responses. This pilot study integrates molecular biology and explainable machine learning to characterize this pathway, emphasizing its negative regulator PTEN , in stratifying disease severity. RT-qPCR gene expression profiles from 83 laboratory-confirmed COVID-19 patients were merged with clinical data of lung involvement in three severity classes (mild: n = 40, moderate: n = 24, severe: n = 19). Molecular analysis revealed significant upregulation of IL-6 , PI3K , mTOR , and AKT1 in severe and fatal cases, alongside progressive downregulation of PTEN ( P < 0.001) and Cyclin D1 , suggesting impaired pathway inhibition and disrupted cell-cycle regulation. MMP-2 levels correlated with lung involvement. Eight machine learning classifiers were evaluated using Stratified 5-Fold Cross-Validation for multiclass severity stratification. Random Forest achieved the highest performance (Accuracy: 0.879 ± 0.041; F1-score: 0.874 ± 0.045; AUC-ROC: 0.978 ± 0.007), with robust per-class discrimination including the Severe class (F1: 0.727). Crucially, SHAP-based explainable AI independently validated the molecular findings by identifying PI3K (mean |SHAP| = 0.0984) as the most discriminative feature, followed by AKT1 (0.0838), IL-6 (0.0537), Cyclin D1 (0.0327), and PTEN (0.0287), confirming the canonical PI3K → Akt → mTOR signaling hierarchy. These findings demonstrate that integrating explainable AI with molecular validation can identify biologically coherent prognostic biomarkers, offering a transparent, hypothesis-generating framework for future targeted immunomodulatory research in severe SARS-CoV-2 infection.

Authors

Institutions

Publication Details

Journal
Scientific Reports
Published
2026-11-19
DOI
https://doi.org/10.1038/s41598-026-80158-0
Primary Topic
COVID-19 Clinical Research Studies
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Machine learning and molecular validation reveal the PI3K/AKT/mTOR inflammatory axis for therapeutic targeting in severe COVID-19

Hiva Alipanah, Alireza Faghani Ghodrat, Seyedeh-Samira Ashrafmansouri, Sahar Samsami et al.
Scientific Reports
COVID-19 Clinical Research Studies
article

Machine learning and molecular validation reveal the PI3K/AKT/mTOR inflammatory axis for therapeutic targeting in severe COVID-19

Hiva Alipanah, Alireza Faghani Ghodrat, Seyedeh-Samira Ashrafmansouri, Sahar Samsami, Ali Ghanbari Asad, Mohammad Mehdi Naghizadeh
article en

Abstract

Dysregulation of the PI3K/AKT/mTOR signaling axis contributes to COVID-19 severity by amplifying hyperinflammatory responses. This pilot study integrates molecular biology and explainable machine learning to characterize this pathway, emphasizing its negative regulator PTEN , in stratifying disease severity. RT-qPCR gene expression profiles from 83 laboratory-confirmed COVID-19 patients were merged with clinical data of lung involvement in three severity classes (mild: n = 40, moderate: n = 24, severe: n = 19). Molecular analysis revealed significant upregulation of IL-6 , PI3K , mTOR , and AKT1 in severe and fatal cases, alongside progressive downregulation of PTEN ( P < 0.001) and Cyclin D1 , suggesting impaired pathway inhibition and disrupted cell-cycle regulation. MMP-2 levels correlated with lung involvement. Eight machine learning classifiers were evaluated using Stratified 5-Fold Cross-Validation for multiclass severity stratification. Random Forest achieved the highest performance (Accuracy: 0.879 ± 0.041; F1-score: 0.874 ± 0.045; AUC-ROC: 0.978 ± 0.007), with robust per-class discrimination including the Severe class (F1: 0.727). Crucially, SHAP-based explainable AI independently validated the molecular findings by identifying PI3K (mean |SHAP| = 0.0984) as the most discriminative feature, followed by AKT1 (0.0838), IL-6 (0.0537), Cyclin D1 (0.0327), and PTEN (0.0287), confirming the canonical PI3K → Akt → mTOR signaling hierarchy. These findings demonstrate that integrating explainable AI with molecular validation can identify biologically coherent prognostic biomarkers, offering a transparent, hypothesis-generating framework for future targeted immunomodulatory research in severe SARS-CoV-2 infection.

Scientific Reports
Shiraz University (IR), Fasa University of Medical Sciences (IR), Fasa University (IR), Iran University of Science and Technology (IR)
COVID-19 Clinical Research Studies
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