Single-Cell RNA-Seq Cell-Type Classification

A denoising autoencoder and a deep classifier (a supervised autoencoder) read the raw UMI counts of single cells from human blood (peripheral blood mononuclear cells, PBMC) and name the type of every cell: B cell, CD14+ monocyte, CD34+ progenitor, CD4+ T cell, CD8+ T cell or NK cell, or "unknown" when the cell looks like none of them. The models are trained on 17,594 cells of ten populations that were purified by cell sorting before sequencing (10x Genomics), with Seurat-style normalisation and 2,000 highly variable genes, and are compared with PCA, logistic regression and kNN. They are then tested on held-out sorted cells and on two other donors: PBMC 3k with CELLxGENE labels and PBMC 68k. A complete stand-alone MATLAB R2026a version is included.

Category: Bioinformatics - Transcriptomics and Gene-Expression Analysis (Deep Learning, Autoencoders, Deep Classifiers, Single-Cell RNA-Seq, MATLAB, Python)

Transcriptomics & Gene-Expression Analysis Project B13 from ProTechSC Academy, with graphical abstract, video demonstration and project guidance.