Year: 2026
Type: dissertation
Abstract: Traditional transcriptomic classification approaches may obscure cell-type-specific disease signals when heterogeneous cellular populations are treated as a single analytical unit. Single-cell RNA sequencing addresses this limitation by providing cellular resolution, but its datasets are typically high-dimensional, sparse, imbalanced, and susceptible to information leakage when feature selection or model optimization is performed outside the training data. This study presents a computational framework for single-cell transcriptomic classification that integrates feature selection, supervised learning, multiobjective model selection, and SHAP-based interpretation. The framework was evaluated on a chronic obstructive pulmonary disease single-cell RNA-sequencing dataset, restricting the analysis to male samples to reduce sex-related transcriptional confounding. After global filtering, the dataset retained 69,366 cells and 31,902 genes, with 27 cell populations eligible for downstream modeling. For each cell population, Wilcoxon-based feature selection was performed within cross-validation folds, and Logistic Regression, Support Vector Machine, Random Forest, K-nearest neighbors, and Multilayer Perceptron classifiers were evaluated using accuracy, sensitivity, specificity, and F1-score. The proposed strategy was also compared with alternative feature-selection methods, including ANOVA, Random Forest importance, and recursive feature elimination. The results showed that predictive performance varied across cell populations, indicating that no classifier was universally optimal. In the CD4-positive, alpha-beta T-cell case study, the Multilayer Perceptron achieved the most balanced metric profile, represented the only Pareto-optimal solution, and maintained comparable performance in the final hold-out evaluation. SHAP-based interpretation highlighted influential genes, while enrichment analysis associated the selected features with translation, stress-response, and regulatory processes.
Author: Nunes, Franklin Vinny Medina
Language: Portuguese
Subfield: Molecular Biology
Domain: Life Sciences
Open Access status: green
OpenAlex ID: W7171605709