Explainable machine learning for predicting lower extremity deep vein thrombosis in traumatic brain injury patients: development of a SHAP-guided Random Forest model.

The accelerated global population aging represents the most prominent demographic shift of the 21st century, posing substantial challenges to public health systems worldwide and socio-economic development. According to statistics from the World Health Organization (WHO), by 2050, the global population aged 60 and above is projected to exceed 2 billion, accounting for over 22% of the total global population (; ). Consequently, the incidence of age-related diseases (such as cardiovascular disease, neurodegenerative diseases, metabolic disorders, and malignant tumors) continues to rise, not only severely affecting the quality of life for the elderly but also imposing a substantial burden on healthcare systems (). Against this backdrop, slowing aging, extending healthspan, and reducing the risk of age-related diseases have emerged as critical challenges and research priorities in the life sciences field (). Aging constitutes a complex biological process involving multiple dimensions, levels, and pathways. According to the updated framework proposed by López-Otín and colleagues (), aging is driven by twelve interconnected hallmarks that fulfill three key criteria: age-associated manifestation, acceleration of aging upon experimental accentuation, and deceleration upon therapeutic intervention. These hallmarks are organized into three categories—primary (genomic instability, telomere attrition, epigenetic alterations, loss of proteostasis, disabled macroautophagy, deregulated nutrient-sensing, mitochondrial dysfunction), antagonistic (cellular senescence), and integrative (stem cell exhaustion, altered intercellular communication, chronic inflammation, dysbiosis). Together, these hallmarks constitute a comprehensive framework for understanding the molecular and cellular basis of aging and for identifying potential therapeutic targets. Traditional anti-aging drug research predominantly focuses on single targets or signaling pathways, rendering it inadequate to comprehensively address the systemic and networked nature of aging. Consequently, most drug development efforts face the dilemma of “limited efficacy and significant side effects” (; ). However, it is important to recognize that currently available anti-aging interventions exhibit distinct and class-specific adverse effect profiles. For example, the mTOR inhibitor rapamycin, while demonstrating lifespan extension in multiple model organisms, is associated with immunosuppression, hyperlipidemia, and insulin resistance in clinical settings (). Senolytic agents, such as the Dasatinib + Quercetin combination or the BCL-2/BCL-xL inhibitor Navitoclax, show efficacy in clearing senescent cells but carry risks including hematological toxicity (thrombocytopenia, neutropenia), fatigue, and gastrointestinal disturbances (; ). Metformin, widely used for and investigated for geroprotective effects, is generally well-tolerated but can cause gastrointestinal intolerance and, rarely, lactic acidosis (). Even aspirin, considered for its anti-inflammatory geroprotective potential, demonstrates a dose-dependent increase in bleeding risk (OR = 1.54, 95% CI 1.32–1.80) (). These observations underscore that no universal ‘safe’ anti-aging agent exists; instead, risk-benefit assessment must be individualized based on therapeutic class, dosing regimen, patient comorbidities, and frailty status. Consequently, most drug development efforts face the dual challenge of achieving sufficient efficacy while managing predictable, class-specific adverse effects. In recent years, multi-omics technologies have generated high-dimensional molecular landscapes that systematically define hallmarks of aging, including genomic instability, mitochondrial dysfunction, cellular senescence, chronic inflammation, and dysbiosis (; ; ). Meanwhile, artificial intelligence provides the analytical power to integrate heterogeneous omics data, identify core aging signatures, prioritize therapeutic targets, and predict candidate anti-aging compounds (; ; ; ; ).