Sivakumar Vengusamy

Background: Alzheimer’s disease (AD) remains difficult to diagnose due to molecular heterogeneity and the limitations of existing biomarkers in invasiveness, cost, and cross-cohort reproducibility. This study aimed to identify reproducible AD-associated candidate diagnostic genes and nominate potential compound leads using an integrative multi-cohort transcriptomic and ensemble machine-learning framework. Methods: Public AD brain transcriptomic datasets were integrated under a predefined leakage-controlled multi-cohort design, with separate training and external validation cohorts. Candidate genes were prioritized through differential expression analysis (DEA), weighted gene coexpression network analysis, and protein-protein interaction (PPI) network ranking, and then evaluated using 127 prespecified machine-learning model-feature combinations with hyperparameters optimized by internal cross-validation (CV). Final hub genes were selected by integrating SHapley Additive exPlanations (SHAP)-derived model contribution, bootstrap stability, PPI-rank stability, and cross-cohort reproducibility. Model interpretability and biological relevance were assessed using SHAP analysis, immune infiltration deconvolution, and single-cell transcriptomic analysis. Candidate compounds were prioritized through drug-gene enrichment, ADMET (absorption, distribution, metabolism, excretion, and toxicity) profiling, and molecular docking. Results: A total of 2,534 differentially expressed genes (DEGs) intersecting with weighted gene co-expression network analysis (WGCNA) modules yielded 848 candidates; PPI prioritization retained 50. The random forest (RF) model achieved an area under the curve (AUC) =0.957 [95% confidence interval (CI): 0.922–0.984], sensitivity 0.978, and specificity 0.936. IGF1R and SPP1 maintained AUC >0.70 across all three external cohorts. Four docking compounds showed binding energies below −9.0 kcal/mol, including one at −10.5 kcal/mol. Conclusions: This framework nominates IGF1R and SPP1 as reproducible AD-associated candidate diagnostic genes and identifies four compound leads, all requiring experimental and clinical validation.

This study examines the effectiveness of Genetic Algorithm (GA)-based image augmentation in enhancing Convolutional Neural Network (CNN) performance for Romblon marble texture classification. Five CNN architectures, AlexNet, InceptionV3, VGG16, MobileNet, and ResNet50, were trained and evaluated on both raw and GA-augmented datasets. Unprocessed dataset achieved limited accuracy, with InceptionV3 performing best at 35.33%, followed closely by VGG16 at 33.78%. In contrast, GA-augmented data significantly boosted performance, with VGG16 achieving 94.68% accuracy, followed by MobileNet (92.68%) and InceptionV3 (92.46%). Entropy loss values consistently decreased across all models, indicating improved convergence and reduced overfitting. Although ROC-AUC scores remained close to 0.5, reflecting modest improvements in class separability, overall results confirm that evolutionary augmentation enriches dataset diversity and strengthens CNN learning capacity. MobileNet showed a solid balance between accuracy and computational economy, underscoring the possibility of GA-based augmentation as a workable option for real-world marble categorization, while VGG16 emerged as the most accurate of the studied architectures.

Open-set recognition of wild plants in natural complex scenes is an important task for plant conservation, ecological monitoring, and precision agriculture. Traditional closed-set learning methods struggle to handle unseen species not covered by the training set and complex environmental interferences, while existing open-vocabulary methods lack knowledge-driven reasoning capabilities and cannot provide interpretable recognition for unknown categories. This research proposes the Reasoning-Aware Perceptual Framework that integrates open-vocabulary vision-language models, foundation mask-generation tools, and domain knowledge reasoning to achieve known/unknown category recognition, online perception, and interpretable reasoning of unknown wild plant species. Centered on a five-stage closed loop of Perception-Retrieval-Reasoning-Decision-Iteration, the framework captures open concepts through vision-language feature alignment, completes evidence-based reasoning and confidence evaluation in combination with a botanical domain knowledge base, and finally outputs species classification decisions, interpretable reasoning reports with family/genus-level taxonomic affinity, and uncertainty-calibrated confidence scores. The unknown category estimation with family/genus-level taxonomic affinity in this framework refers to a general unknown label combined with taxonomic affinity at the family/genus level, which can clearly reflect the evolutionary relationship between unknown species and known species. Experiments on the self-constructed WildPlantOpenSet-10K dataset and public benchmark datasets report an F1-score of 84.7% for unknown species recognition, AUROC of 0.93 for known/unknown discriminability, and mean F1 of 87.0% across all categories. This framework focuses on open-set wild plant recognition and interpretable reasoning, using off-the-shelf instance extraction to acquire visual features for downstream reasoning. It maintains stable robustness in complex scenarios such as occlusion, strong light, and multi-species coexistence, and can adapt to the open-world environment without relying on large-scale pixel annotations, providing a research prototype for interpretable open-set recognition in complex natural environments.

Background: Breast cancer (BC) is one of the most diagnosed malignancies and a leading cause of cancer-related mortality among women worldwide, thereby posing a substantial threat to women’s health worldwide. However, clinically robust diagnostic biomarkers with high sensitivity and specificity, as well as well-validated molecular targets for targeted therapy, remain limited. Methods: BC transcriptomic data from seven GEO datasets and the TCGA-BRCA cohort (n = 1231) were integrated for analysis. After batch-effect correction, candidate genes were screened through DEA, WGCNA, and PPI networks analysis. An ensemble machine learning (ML) framework incorporating 127 algorithmic combinations was constructed, and SHAP analysis was applied to identify hub genes. Further analyses included functional enrichment, immune infiltration, miRNA regulatory network analysis, and SMR analysis. The expression patterns were validated using single-cell transcriptome data. Drug repositioning analysis and AI-assisted virtual screening were performed to prioritize compounds with favorable drug-like properties. The predicted binding modes of candidate compounds with CHEK1 were assessed by molecular docking. Results: Thirty core genes were obtained through differential expression, WGCNA, and PPI screening. Integrated ML (127 algorithms) determined the optimal model (AUC = 0.919), and SHAP identified nine feature genes, among which CHEK1 and KIF23 showed preliminary diagnostic potential across four external cohorts (AUC: 0.625–0.938). Functional enrichment indicated that both are enriched in the cell cycle and p53 pathways, closely associated with BRCA1/ATR; immune infiltration revealed significant correlations with macrophages and CD8⁺ T cells, with hsa-miR-15a-5p and hsa-miR-607 being common upstream regulatory miRNAs. SMR analysis supported a causal relationship between CHEK1 expression and BC genetic susceptibility (p_SMR < 0.05, p_HEIDI > 0.05); single-cell analysis confirms its heterogeneous expression. AI-assisted virtual screening identified 25 A-grade computational candidate compounds from 171 candidates. Molecular docking suggested that Olaparib and LY294002 can form favorable interactions with the CHEK1 active pocket. Conclusions: The study identified CHEK1 as a key diagnostic gene for BC through 127 ML algorithms and SMR causal inference. By combining AI-assisted virtual screening and molecular docking, computational candidate compounds targeting CHEK1 were prioritized. These findings represent hypothesis-generating in silico predictions and require experimental validation before any therapeutic conclusions can be drawn.