Related Experiment Video
Updated: Jun 12, 2026
10:24
SILAC Based Proteomic Characterization of Exosomes from HIV-1 Infected Cells
Published on: March 3, 2017
Machine-Learning Approach to Identify Tissue Inhibitors of Metalloproteinases (TIMP) and Clinical Variables
Ya-Wei Weng1,2, Hung-Chin Tsai2,3,4,5, Susan Shin-Jung Lee2,3,4
1Institute of Clinical Medicine, College of Medicine, National Cheng Kung University, Tainan, Taiwan.
Open Forum Infectious Diseases
|June 11, 2026
Summary
Executive dysfunction is common in people with HIV (PLWH). Machine learning models integrating clinical data and TIMP levels accurately predict impairment, aiding early intervention for HIV-associated neurocognitive disorder.
Area of Science:
- Neuroscience
- Immunology
- Biochemistry
Background:
- Executive function impairment is prevalent in people living with HIV (PLWH) despite effective combination antiretroviral therapy (cART).
- HIV-associated neurocognitive disorder pathogenesis is multifactorial, involving immune activation, metabolic changes, and neuroinflammation.
- The role of matrix metalloproteinases (MMPs) and their tissue inhibitors (TIMPs) in HIV-related executive dysfunction and blood-brain barrier integrity is not well understood.
Purpose of the Study:
- To investigate the relationship between TIMPs and executive function in middle-aged PLWH.
- To develop and evaluate machine learning models for predicting executive dysfunction in PLWH.
- To identify key clinical and biological predictors of executive dysfunction.
Main Methods:
- 169 middle-aged Taiwanese PLWH were assessed for executive function using the Wisconsin Card Sorting Test.
- Predictive models were developed using logistic regression, support vector machine, random forest, and XGBoost, incorporating variables like years since HIV diagnosis, BMI, nadir CD4 count, and plasma TIMP levels (TIMP-1, TIMP-2, TIMP-4).
- Model performance was evaluated using area under the receiver operating characteristic curve (AUC) with 10-fold cross-validation and SHAP analysis for feature interpretation.
Main Results:
- Executive impairment was present in 9.5% of participants.
- TIMP-1 showed the highest predictive performance as a single predictor (AUC = 0.76 by XGBoost).
- Combining clinical and laboratory variables significantly improved prediction, with XGBoost achieving an AUC of 0.89; cross-validation confirmed model robustness (mean AUC = 0.77).
Conclusions:
- Executive dysfunction is a significant neurocognitive issue in middle-aged PLWH.
- Machine learning models integrating clinical and biological data, particularly TIMP levels, can effectively identify individuals at risk for executive dysfunction.
- Early identification supports timely neuropsychological evaluation and intervention for HIV-associated neurocognitive disorder.