Analyze
ExplainChoose an analysis workflow
Models, single-sample healthy atlas comparison, and multimodal RNA/ATAC/methylation analysis are now grouped here.
Research only: Outputs are exploratory and are not medical advice, diagnosis, or treatment.
Healthy Atlas Compare
Compare one sample against a healthy reference atlas and review deviation, pathway, and top-gene signals.
Use belowMultimodal VAE
Train RNA, ATAC, and methylation together in a shared latent space with disagreement/anomaly outputs.
Open Multimodal ↑Multimodal Shared Latent Model
Aligns RNA, ATAC, and methylation matrices by sample ID, preprocesses each modality, builds a shared latent biological state space, and reports per-modality reconstruction error, disagreement, and research anomaly scores.
Current MVP uses a lightweight shared latent engine; the PyTorch PoE VAE scaffold is included for the next training-service phase.
Health Atlas Comparator
Compares a personal expression sample against a healthy baseline atlas to estimate how far the sample deviates from normal reference biology.
Disease Pattern Similarity
Compares the abnormal portion of a sample against curated biological signatures such as interferon-high viral inflammation, autoimmune-like activation, neutrophil-dominant stress, or metabolic inflammatory shift.
This is a resemblance engine, not a diagnosis engine.
Healthy Atlas Compare
Upload your reference atlas and personal sample
Healthy atlas format: rows = healthy reference samples, columns = genes, values = counts or normalized expression. Personal sample format: either one row with genes as columns, or two columns named gene_id and value.