Methods are named so you can judge them. Every pipeline runs with Padma in the workspace on the same sealed engines, and every one is also available as a scientist-led project. The engines and the references behind them are updated continuously.
scanpy / scvi-tools pipeline with donor-level statistics enforced by the QC gate.
Methods · MAD-based QC; doublet detection; scVI/scANVI integration; annotation; pseudobulk differential expression by donor (no pseudoreplication); trajectory
Differential expression and pathway analysis with the experimental unit and batch structure declared up front.
Methods · Alignment and quantification; DESeq2 / edgeR; fgsea / GSEA over MSigDB (H, C2, C3, C5, C6, C7); deconvolution
Joint RNA and surface-protein analysis with totalVI.
Methods · totalVI on raw counts; joint protein–RNA analysis; surface-marker validation of transcriptional clusters
Somatic and germline variant analysis on community-validated workflows.
Methods · Somatic SNV/indel/CNV/SV (tumour–normal and tumour-only); germline; nf-core/sarek where applicable
Chromatin accessibility alone or paired with RNA.
Methods · PeakVI, snapATAC2, MultiVI; motif and gene-regulatory-network inference
Genotyping and clone calling from single-cell DNA; feeds the clonal-evolution module.
Methods · Mission Bio Tapestri genotyping; clone calling; co-occurrence
Spatial deconvolution and niche analysis.
Methods · Deconvolution (DestVI); niche analysis; cell–cell interaction
Panel variant calling and annotation; feeds the CH and clonal-evolution modules.
Methods · Variant calling; interpretation against ACMG/AMP, OncoKB, COSMIC, gnomAD; TP53 mono- vs bi-allelic state
Array and sequencing-based methylation analysis.
Methods · EPIC array; EM-seq / RRBS; DMR calling
Isoform, SV and phasing analysis from long reads.
Methods · ONT / PacBio isoform analysis; structural variants; phasing
Repertoire analysis and clonotype tracking.
Methods · BCR/TCR clonotyping; diversity; clonotype tracking (shared machinery with MRD)
Integration across modalities and biomarker models with cross-validation that holds up.
Methods · MOFA-style factor models; supervised biomarker models with nested cross-validation
Population discovery and donor-level differential abundance from spectral flow and CyTOF files.
Methods · FCS 2.0/3.0/3.1 intake; compensation; arcsinh / logicle transforms; FlowSOM-style clustering; differential abundance and state at donor level (diffcyt-style)
Protein abundance and immunoassay quantification with the experimental unit declared up front.
Methods · Mass spectrometry, Olink and SomaScan abundance (limma / MSstats, imputation sensitivity); ELISA and immunoassay quantification to ICH M10 conventions
Image-derived features and their statistics, from raw microscopy to a comparison that respects the unit.
Methods · Microscopy, IF and IHC quantification; nuclei and cell segmentation; morphometric features; group comparisons at donor level
Dose response, synergy and plate screens with replicate structure and plate effects modelled.
Methods · Dose–response curves (IC50 / AUC, four-parameter logistic); combination synergy (Bliss, Loewe, HSA); plate-screen normalisation and hit calling
Relative expression from RT-qPCR with reference-gene checks and MIQE reporting.
Methods · ΔΔCt relative expression; reference-gene stability (geNorm-style); efficiency correction; MIQE-conformant reporting
Outcomes, risk scores and treatment effects on clinical-research tables, with the assumptions written down.
Methods · Kaplan–Meier and Cox models; competing risks; risk-score validation (calibration, discrimination); causal treatment effects (DAG backdoor check, doubly robust estimation, E-values)
GWAS and polygenic scores with QC and structure handled before any association is reported.
Methods · Genotype QC; association testing with population structure; polygenic score construction and validation
Measurable-residual-disease analysis from NGS with an explicit limit of detection on every report. Research-grade; never a clinical result.
Methods · Error-corrected deep sequencing (UMI/duplex); variant tracking across timepoints; IG/TCR clonotype tracking; low-VAF calling; sensitivity and limit of detection reported per assay and sample
Clonal architecture and how it changes under treatment, from bulk VAF or single-cell DNA.
Methods · Bulk VAF and single-cell DNA (Tapestri) clone reconstruction; phylogenetic trees; timepoint dynamics; therapy-pressure analysis
CH calling, VAF trajectories and risk annotation from panels, WES or WGS, in any cohort where CH is a covariate or an endpoint.
Methods · CH variant calling; VAF trajectories; annotation against gnomAD, OncoKB, COSMIC; TP53 allelic state
Linking genotype, transcriptome and surface phenotype across modalities measured on the same individual.
Methods · scRNA + scDNA + CITE-seq + ATAC from one individual; genotype-to-phenotype linking of clones; MOFA-style factor models
Every pipeline ships with
A versioned container; a pinned runner; the QC gates (sample QC, pseudoreplication, batch–outcome confounding, multiple testing, power, missing controls); a provenance record per run; figures, tables and public-repository submission support.
