SoftMoE — multi-omics imputation
Reconstruct a missing omics layer (CNV, methylation, miRNA, RNA-seq) from the others with a Soft Mixture-of-Experts model trained on ISB-CGC (TCGA, TARGET, CPTAC, …). Upload CSVs in raw units (rows = samples with ID in the first column, columns = feature names; methylation as beta-values, CNV as copy number — its z-score scale is only approximated, mean 3.1 / std 1.8). Inputs are z-scored with the training-set mean/std and imputed layers are returned in raw units. Leave the missing layer empty or select it below. Example CSVs are in the examples folder of this Space.
Research prototype — not for clinical use.
Preview of first imputed layer (10 samples × 12 features)
Example: impute RNA-seq from CNV + methylation + miRNA
| CNV CSV | Methylation CSV | miRNA CSV | Force these to be treated as missing (optional) | Input is already z-scored (skip normalisation; output stays z-scored) |
|---|