Scientists from Harbin Medical University and Nantong Tumor Hospital develop integrative machine learning models combining gene co-expression analysis and multi-omics data to predict prostate cancer diagnosis and biochemical recurrence risk, enhancing personalized precision oncology.

Key points

• Applied WGCNA to identify 16 BCR-related genes and correlated modules with clinical recurrence outcomes in TCGA-PRAD.
• Constructed LASSO+LDA diagnostic model validated across five independent cohorts, achieving AUCs up to 0.897.
• Used XGBoost and SHAP analyses to pinpoint COMP as a high-impact biomarker and validated its functional role via molecular docking and in vivo assays.

Why it matters: Integrating machine learning with gene expression profiling enhances precision oncology by improving early detection and individualized recurrence risk assessment beyond conventional methods.

In this overview article, Melissa D. Johnson surveys the modern anti-aging movement by examining telomeres—protective DNA endcaps—as fundamental regulators of cellular senescence and immortality. She traces historical milestones from free radical theory to life extension societies, discussing how telomere maintenance and manipulation may underlie future therapies for aging-related diseases and cancer prevention.

Key points

• Telomeres act as protective DNA endcaps regulating cellular lifespan and senescence.
• Critically short or dysfunctional telomeres trigger replicative senescence or chromosomal instability leading to cancer.
• Research focuses on telomerase activation and advanced biotechnologies to repair telomeric damage and extend healthspan.

A team at USC’s Keck School of Medicine applies demixed PCA to high-gamma SEEG signals from the insular cortex, then trains a bidirectional LSTM network to classify left, right, or rest movements. They achieve 73% accuracy, significantly above chance, offering new deep-structure signals for motor BCIs.

Key points

• Insular SEEG recordings focus on high-gamma band (70–200 Hz) activity during left, right, and no-movement trials.
• Demixed PCA extracts ten stimulus-dependent dimensions that separate movement conditions in latent space.
• Bidirectional LSTM network decodes movement direction with 72.6% ± 13.0% accuracy, surpassing 33.3% chance level.

Why it matters: Demonstrating robust directional decoding from insular high-gamma signals opens deep-brain sources for more precise, distributed motor BCIs.