eHSCPr

eHSCPr predicts early hematopoietic stem cell (HSC) developmental stages from single-cell transcriptomic data using machine learning.

Key Features:

• Input data: Processes large-scale single-cell transcriptomic datasets.
• Gene selection (F-score comparison): Uses the F-score for differential gene selection and compares it with limma, DESeq2, and edgeR, reporting an ROC AUC of 0.987 for selected markers.
• Surface marker identification: Identifies critical surface markers for endothelial cells and hematopoietic cells.
• Classifier: Implements a support vector machine (SVM)-based classifier.
• Validation metrics: Reports 10-fold cross-validation accuracies of 94.19% on the training dataset and 94.84% on an independent dataset, alongside ROC AUC values.
• Transcription analysis: Performs transcription analysis on the F-score gene set to enrich signal markers associated with HSC developmental stages.

Scientific Applications:

• Early HSC stage prediction: Predicts early developmental stages of hematopoietic stem cells from single-cell data.
• Surface marker selection: Selects candidate surface markers for endothelial and hematopoietic cell populations.
• In vitro blood regeneration: Provides marker and stage predictions to inform in vitro blood regeneration studies.
• Extracorporeal blood research: Supports extracorporeal blood research by identifying early HSC developmental signatures.
• Experimental design support: Supports hypothesis-driven experimental design through computational marker and stage predictions.

Methodology:

Processes single-cell transcriptomic data with machine learning using F-score-based differential gene selection compared against limma, DESeq2, and edgeR, applies a support vector machine (SVM) classifier with 10-fold cross-validation, and conducts transcription analysis on the F-score gene set.