Hasty Briefs - Daily Tech News Brief

Bilingual

Development and validation of a prediction model for long-term cognitive frailty risk in stroke patients based on CHARLS data - PubMed
a month ago
- Study developed ML models to predict long-term cognitive frailty risk in stroke patients using CHARLS data.
- XGBoost and Random Forest models showed highest performance (AUC 0.810 and 0.795).
- Key predictors identified: age, education, nutritional status, physical exercise, and IADL.
- SHAP values highlighted age and education as most significant factors.
- Model aims for early screening in primary care and targeted interventions.
Integration of Machine Learning With PBPK and QSAR Modeling Approaches to Facilitate Drug Discovery and Development - PubMed
a month ago
- This review explores how machine learning (ML) improves physiologically based pharmacokinetic (PBPK) modeling by enhancing quantitative structure-activity relationship (QSAR) predictions for drug ADME properties.
- Traditional PBPK models are limited by unreliable input parameters for new compounds, but ML-enhanced QSAR models predict these parameters directly from chemical structures, expanding PBPK applicability.
- Recent advances in ML-powered QSAR lead to better predictions of parameters like partition coefficients and clearance rates, improving PBPK simulation accuracy when integrated as inputs.
- This integration supports drug discovery and development decisions, from candidate selection to safety assessment, potentially reducing pharmacokinetic-related failures.
- Challenges include limited chemical diversity in training data, model interpretability, and regulatory considerations for ML-derived parameters in PBPK submissions.
Machine learning predicts sepsis deterioration trajectories - PubMed
a month ago
- Sepsis has heterogeneous clinical trajectories, and dynamic prediction is crucial for personalized intervention.
- An ensemble machine-learning model identified three latent recovery patterns: rapid recovery (41.5%), slow recovery (36.4%), and clinical deterioration (22.1%).
- The model showed strong performance with AUROC up to 0.92 in development and maintained accuracy in external validation across multiple datasets.
- Reduced heart rate variability (SD < 10 bpm) was a significant predictor of mortality, with an adjusted hazard ratio of 2.17.
- Implementation of the model led to clinical improvements: reduced ICU stay by 1.8 days, mechanical ventilation by 2.3 days, and 28-day mortality by 5.7%.
- The study provided early warnings for deterioration with a median lead time of 17.6 hours, enabling proactive care strategies.
Machine learning analysis of population-wide plasma proteins identifies hormonal biomarkers of Parkinson's disease - PubMed
a month ago
- The study uses machine learning on plasma proteomics from 43,408 UK Biobank subjects to identify Parkinson's disease biomarkers.
- Identified biomarkers include known markers DDC and CALB2, and new markers linked to JAK-STAT, PI3K-AKT pathways, and hormonal signaling.
- Biomarkers correlate with disease severity (UPDRS scores) and are classified as protective or risk-associated, aiding in patient stratification and therapy development.
Assessment of the risk of bacteremia in patients with hematologic malignancies in the emergency department: A comparative study between logistic regression and machine learning algorithms - PubMed
a month ago
- The study compared logistic regression and machine learning algorithms for predicting bacteremia risk in hematologic malignancy patients in the emergency department.
- Three models were tested: BOSS-1 (multivariable logistic regression), BOSS-2 (K-means clustering), and BOSS-3 (support vector machine).
- BOSS-1 had high sensitivity (94%) for low-risk patients but low specificity (30%), while BOSS-2 improved high-risk identification with similar specificity.
- BOSS-3 classified patients into only low-risk (66.8%) or high-risk (33.2%) groups, showing the best balance of sensitivity (61%) and specificity (71%).
- External validation found BOSS-3 had the most reproducible results, indicating supervised machine learning has the greatest clinical potential.

Hasty Briefsbeta

#machine learning

Development and validation of a prediction model for long-term cognitive frailty risk in stroke patients based on CHARLS data - PubMed

Integration of Machine Learning With PBPK and QSAR Modeling Approaches to Facilitate Drug Discovery and Development - PubMed

Machine learning predicts sepsis deterioration trajectories - PubMed

Machine learning analysis of population-wide plasma proteins identifies hormonal biomarkers of Parkinson's disease - PubMed

Assessment of the risk of bacteremia in patients with hematologic malignancies in the emergency department: A comparative study between logistic regression and machine learning algorithms - PubMed