ALS Cross-Platform Classification - CyTOF and scRNA-seq Data

This dataset contains mass cytometry (CyTOF) and single-cell RNA sequencing (scRNA-seq) data from peripheral blood mononuclear cells (PBMCs) of ALS patients and healthy controls, used for developing cross-platform machine learning models to classify disease progression rates. Dataset 1: als_cytof_data.h5ad Technology: Mass cytometry (CyTOF) Samples: 44 samples across 4 batches Groups: Healthy controls, ALS patients (slow, standard, and fast progression) Cell types: 25 immune cell populations profiled Contains: Single-cell protein expression, cell type annotations, clinical metadata (progression rate ALSFRS-R/time, progression group) Dataset 2: Itou2024_scrna_data.h5ad Technology: 10x Genomics scRNA-seq Source: Itou et al. 2024 (GSE244263) Samples: 40 PBMC samples (30 ALS, 10 controls) Cells: 29,883 (750 per donor, subsampled with random_state=42) Genes: 3,054 (highly variable genes only) Groups: Healthy controls, rapid ALS, non-rapid ALS (column: Classifier) Contains: Log-normalized expression (X), cell type annotations, Leiden clusters, PCA/Harmony/UMAP embeddings Use case: Cross-platform classification analyses (companion to CyTOF data) Dataset 3: Itou2024_scrna_data_full_with_counts.h5ad (new) Technology: 10x Genomics scRNA-seq Source: Itou et al. 2024 (GSE244263) — same cells as Dataset 2 Samples: 40 PBMC samples (30 ALS, 10 controls) Cells: 29,883 (750 per donor, subsampled with random_state=42) Genes: 36,601 (full transcriptome) Groups: Healthy controls, rapid ALS, non-rapid ALS Contains: X: Log-normalized expression (sparse, float64) layers['counts']: Raw integer counts (sparse, int32) Cell type annotations (26 cell types), harmonized labels, ALS subtype metadata PCA, Harmony-corrected PCA, and UMAP embeddings Use case: Differential expression (DESeq2/edgeR require raw counts), cell-cell communication analysis, and any reanalysis requiring the full gene set Note: This is the same set of 29,883 cells as Dataset 2, expanded to the full transcriptome with raw counts preserved. Dataset 2 contains only highly variable genes and is sufficient for the cross-platform classification analyses described in the manuscript. These datasets enable cross-platform validation of immune signatures in ALS, demonstrating that disease-associated immune patterns identified in CyTOF data generalize to scRNA-seq data. The datasets support reproducible machine learning analyses for ALS patient stratification based on immune profiling. Manuscript: Peripheral immune patterns enable robust cross-platform prediction of ALS onset and progression