All recipes

Every catalogued recipe, one page each. Use the sidebar (or the search box at the top of the page) to jump to a specific recipe. To browse by problem class or see where the gaps are, see the Landscape. To return to the section overview, see Recipes.

---

Table of contents

• Analyze an existing MD trajectory for stability, flexibility, and contacts - Drive the MDAnalysis skill (MDTraj as backup) to take a finished GROMACS/AMBER/NAMD trajectory through RMSD/RMSF/Rg, contact maps, H-bonds, and PCA without writing the analysis script by hand.
• Analyze a Perturb-seq CRISPR screen for perturbation effects - Use the Perturb-seq Analysis skill to assign guides, remove escapers, test with a calibrated method, and rank perturbations by E-distance.
• Analyze the SAR of a measured compound series - Drive the SAR Analysis skill over a CSV of assayed analogs to decompose the common scaffold, tabulate R-group vs potency, and surface activity cliffs for the next lead-optimization round.
• Analyze a single-cell TCR repertoire alongside gene expression - Use the scirpy Analysis skill to QC paired single-cell TCR chains, define clonotypes, quantify clonal expansion, and overlay clonality on the transcriptomic UMAP.
• Annotate TCR antigen specificity by clustering and database lookup - Use the TCR-Epitope Binding skill to group CDR3 sequences into shared-specificity clusters and match them against VDJdb, IEDB, and McPAS-TCR.
• Assemble a tissue reference atlas from the CELLxGENE Census - Use the cellxgene-census Claude skill to slice a tissue- or disease-specific AnnData from the CZ CELLxGENE Census, optionally fetching the precomputed scVI embedding for reference mapping or downstream integration.
• Build a clonal lineage tree from CRISPR scar recorders - Use the Lineage Tracing skill to turn recorder reads into a character matrix, reconstruct trees with several solvers, and check topology robustness before interpreting.
• Annotate a single bacterial genome assembly - Use the Bakta Claude skill to annotate one bacterial or archaeal assembly — CDS, rRNA/tRNA, CRISPR arrays — into NCBI-compatible GFF3/GenBank with a feature summary.
• Annotate cell types in a single-cell dataset - Use the SciAgent CellTypist skill to put reference-backed cell-type labels on a QC'd AnnData, escalating to the popV consensus skill when you need ensemble uncertainty.
• Assemble a public cancer imaging cohort from NCI Imaging Data Commons - Use the Imaging Data Commons skill to query NCI IDC by modality, body part, and collection, then download a reproducible DICOM cohort ready for segmentation.
• Benchmark an ADMET property with PyTDC - Drive the PyTDC Claude skill to load a Therapeutics Data Commons ADMET dataset with its leaderboard split, train or score a baseline model, and emit the standard TDC metrics so a new method can be compared head-to-head against the leaderboard.
• Build a resting-state functional-connectivity matrix from preprocessed fMRI - Use the Nilearn skill in Claude Code to turn fMRIPrep-preprocessed BOLD into an atlas ROI-to-ROI functional-connectivity matrix with confound regression, as a re-runnable script.
• Annotate and verify an engineered plasmid construct - Use the pLannotate skill to auto-annotate an engineered plasmid — promoters, resistance markers, origins, tags — and verify a construct's sequence before you clone or order it.
• Build a pharmacogenomic dosing report from a patient's diplotypes - Turn a patient's PGx diplotypes plus a medication list into CPIC/DPWG genotype-to-dosing recommendations with cited guideline sources.
• Build a phylogenetic tree from a set of sequences - Use the phylogenetics Claude skill to take a FASTA of homologous sequences through MAFFT alignment, IQ-TREE 2 maximum-likelihood inference with bootstrap support, and an annotated tree figure.
• Build a protein interaction network and rank hub genes - Feed a gene list to the STRING skill to build a PPI network, then rank hub proteins by degree/centrality and detect modules — the canonical "which of these genes matter most" step.
• Build a target dossier from gene name to structure to cancer dependency - Combine the Open Targets plugin, UniProt MCP, AlphaFold MCP, and the DepMap skill to produce a one-page dossier on a candidate gene — disease evidence, protein annotation, predicted structure, and cancer-cell-line essentiality.
• Characterize a protein–protein interface from a structure you already have - Use the Interface Analysis skill to get contact residues and buried surface area from a solved or predicted complex — and to decide whether the interface is real or crystal packing.
• Annotate tumor somatic variants with clinical actionability evidence - Use the CIViC connector to match a tumor's somatic variant list to predictive/prognostic/diagnostic clinical evidence and produce an AMP/ASCO/CAP-tiered, cited actionability table for a molecular tumor board.
• Call peaks and find enriched motifs from ChIP-seq or ATAC-seq - Chain the MACS3 peak-calling skill into the HOMER motif skill to turn aligned ChIP-seq/ATAC-seq BAMs into a called-peak set with nearest-gene annotation and enriched TF motifs.
• Compute 16S microbiome alpha/beta diversity from a BIOM table - Use the scikit-bio Claude skill to take a BIOM feature table and sample metadata through Shannon/Simpson/Faith's PD, UniFrac, PCoA, and PERMANOVA in one chat.
• Compute a bacterial pan-genome from a set of genome assemblies - Annotate a panel of bacterial assemblies with Bakta, then cluster genes into core/accessory partitions with Roary to get a pan-genome and presence/absence matrix.
• Compute DTI scalar maps (FA/MD/AD/RD) from diffusion MRI - Use the DIPY skill in Claude Code to fit a diffusion tensor to a DWI volume and emit FA/MD/AD/RD maps plus ROI statistics, as a re-runnable script.
• Compute HRV from an ECG recording - Use the NeuroKit2 Claude Skill to clean an ECG trace, detect R-peaks, and return time-domain, frequency-domain, and non-linear HRV indices from a single conversation.
• Compare bulk TCR/BCR repertoires across samples from raw immunosequencing reads - Assemble clonotypes with MiXCR, normalize samples to a common depth with VDJtools, then compare diversity, clonality and overlap as a committed pipeline.
• Convert raw analytical instrument data to Allotrope ASM JSON - Drop a raw instrument file (cell counter, plate reader, HPLC, qPCR) onto Claude Code and get back valid Allotrope Simple Model JSON, a flattened CSV for LIMS, and a standalone Python parser script.
• Decide whether a suspicious cluster is a doublet artifact - Use the bioSkills Doublet Detection skill plus two orthogonal checks to rule on whether an intermediate single-cell cluster is real or two cells in one droplet.
• Discover NWB recordings on DANDI and prepare them for sorting - Use the Neurosift Tools MCP to find Neuropixels (or other extracellular) NWB sessions on DANDI matching your hypothesis, inspect them in-place, and stage a download list for downstream spike sorting.
• Design a Bayesian trial that borrows external control data - Use the Bayesian Trials skill and RBesT to turn historical controls into a MAP prior with a declared effective sample size, then simulate what it costs under prior-data conflict.
• Design a group-sequential trial with interim looks - Use the Adaptive Designs skill and rpact to fix boundaries, simulate operating characteristics, and pre-specify any sample-size re-estimation rule before enrolment.
• Design CRISPR sgRNAs for a gene knockout - Use the sgRNA Design skill to pick CRISPR-Cas9 knockout guides for a gene — preferring validated/CRISPick guides, falling back to Rule Set 2 + CFD de novo rules.
• Dock a ligand library into a target structure with DiffDock - Drive DiffDock from Claude Code to generate per-ligand binding poses against a PDB or AlphaFold target, filter by confidence, and emit a ranked pose set ready for MM/PBSA rescoring or visualization.
• Design amino-acid sequences for a fixed protein backbone - Design sequences for a target backbone with ProteinMPNN, then refold each candidate with ESMFold and keep only those whose self-consistency RMSD recovers the backbone.
• Draft a Phase 2/3 clinical-trial protocol from an indication brief - Use the Anthropic clinical-trial-protocol plugin to expand a short indication / endpoint brief into an FDA/NIH-compliant Phase 2/3 protocol draft, with regulatory pathway, competitive landscape, and sample-size calculation already wired in.
• Enumerate analogs around a lead compound for SAR expansion - Use the Datamol skill in Claude Code to enumerate standardized, drug-like analogs around a lead SMILES — tautomers, stereoisomers, and fragment-substituted variants — as a ranked SAR-expansion table.
• Estimate pharmacokinetic properties of a small molecule - Combine RDKit physchem descriptors, MedChem rule-based ADMET flags, and measured ChEMBL bioactivity / DMPK endpoints to build a defensible PK estimate for one compound without an ML predictor in the loop.
• Extract event-related potentials from EEG epochs - Use the MNE-Python EEG skill in Claude Code to filter, ICA-clean, epoch around event markers, and average a raw continuous EEG file into per-condition evoked responses.
• Extract structured variables from free-text clinical notes - Use Claude Code alone to extract predefined variables from de-identified clinical notes into a reviewable, provenance-tracked registry table.
• Extract neuronal traces from a calcium imaging movie - Use the Calcium Imaging Analysis Guide skill to pick motion correction, ROI extraction, neuropil correction and dF/F for two-photon or miniscope data, captured as a re-runnable script.
• Extract spectral features from resting-state EEG - Use the EEG Processing skill in Claude Code plus specparam/FOOOF to turn eyes-closed resting recordings into per-channel band power and aperiodic exponents.
• Filter a virtual screening hit list with drug-likeness rules and structural alerts - Use the MedChem skill in Claude Code to cascade Lipinski / Veber / PAINS / BRENK filters over a SMILES hit list and emit a triaged CSV with per-rule flags and a final keep/drop column.
• Fit a dose-response curve and report a defensible IC50 - Drive the ToolUniverse Dose-Response skill to fit 4PL curves to plate-assay data, gate them on fit quality, and emit IC50 with AUC and per-replicate spread.
• Find a selective tool compound for a target - Rank candidate tool compounds for one target by curated affinity and measured selectivity using the GtoPdb skill, with an explicit "not profiled enough to call" verdict.
• Find differentially abundant proteins in a label-free proteomics experiment - Use the MaxQuant skill to take LFQ raw files to a filtered protein table and a differential-abundance list whose imputation sensitivity is reported, not hidden.
• Find differentially regulated phosphosites, not just changed proteins - Use the MaxQuant skill on a paired phospho-enriched and unenriched run to get site-level regulation calls that are not just protein-abundance changes in disguise.
• Find drug-repurposing candidates by walking a biomedical knowledge graph - Walk PrimeKG offline from a disease through its genes to the drugs that target them, then flag drugs whose current indications are elsewhere as repurposing leads.
• Find selective genetic dependencies for a cancer context with DepMap - Use the DepMap skill to mine genome-wide CRISPR screens for genes selectively essential in a cancer lineage or genotype, corrected for copy-number bias.
• Find which taxa differ between microbiome groups - Use the bioSkills Differential Abundance skill to test an amplicon feature table with several compositionally-aware methods and report their consensus, not one tool's hit list.
• Fit a drift-diffusion model to choice and reaction-time data - Use the Drift-Diffusion Model skill in Claude Code with PyDDM to decompose two-choice accuracy and RT into drift rate, boundary, bias and non-decision time.
• Identify a bacterial isolate from its 16S rRNA sequence - Use the BLAST MCP server to search a Sanger 16S rRNA read against a pinned reference database and assign genus/species by the standard identity thresholds.
• Identify an unknown compound from an MS/MS spectrum - Use the matchms skill in Claude Code to clean an unknown MS/MS spectrum and rank candidate identities by spectral similarity against a reference library.
• Infer cell-cell communication from single-cell RNA-seq - Use the LIANA-MCP server to run multiple ligand-receptor methods on an annotated AnnData object, aggregate their ranks, and plot the consensus communication network.
• Infer a gene-regulatory network from single-cell RNA-seq - Use the Arboreto Claude skill to run GRNBoost2 on a QC'd single-cell AnnData and recover a ranked TF–target edge list — the first step of a SCENIC regulon pipeline.
• Harmonize free-text clinical terms to standard codes - Use the Medical Terminologies MCP to map a column of free-text diagnoses, drugs, or labs to ICD-11/SNOMED/RxNorm/LOINC codes as a committed, reviewable crosswalk.
• Integrate multiple single-cell RNA-seq datasets across batches - Use the scvi-tools skill in Claude Code to fit scVI (or scANVI when labels are available) on a concatenated AnnData of multiple batches, returning a batch-corrected latent space and integrated UMAP.
• Map a disease to its implicated genes and pathways - Chain Open Targets target ranking into gget/Enrichr functional enrichment to turn a disease name or EFO/MONDO ID into a cited genes-and-pathways readout.
• Integrate multi-omics layers into interpretable factors with MOFA+ - Use the MOFA+ Claude skill to jointly decompose several omics layers on the same samples into a handful of latent factors, then read off which factors capture biology and which layers drive them.
• Interpret a clinical variant from a natural-language query - Use BioMCP to convert a free-text variant query ("BRAF V600E", "rs113993960", "NM_004985.5:c.35G>A") into a one-page clinical report with ClinVar significance, population frequency, in-silico predictions, and linked literature.
• Infer the function of an uncharacterized protein from its 3D structure - Submit a protein structure to Foldseek, pull structurally similar annotated hits across AFDB/PDB/SwissProt, and infer likely function where sequence search finds nothing.
• Interpret variants that gain or lose glycosylation sites - Flag coding variants that destroy or create N-/O-glycosylation sites with the GlyGen MCP, join to ClinVar/AlphaMissense via BioMCP, and emit a ranked, provenance-tracked candidate table.
• Look up the curated composition and stoichiometry of a protein complex - Have Claude Code script the Complex Portal release files so a gene or UniProt accession returns curated subunits, stoichiometry, and an evidence class.
• Match a patient summary to recruiting clinical trials - Use BioMCP (or the standalone ClinicalTrials.gov MCP) to take a free-text patient summary and return a ranked list of currently-recruiting trials with eligibility rationale.
• Measure cortical thickness and subcortical volumes across a cohort - Drive FreeSurfer recon-all over a T1w cohort with one pinned version, Euler-number QC, and a morphometry table that carries its quality covariate.
• Infer transcription-factor and pathway activities from expression - Drive the decoupler-MCP server to score per-cell or per-sample transcription-factor (CollecTRI) and pathway (PROGENy) activities from an expression matrix, then read off the regulators that move between conditions.
• Localize a fiber/probe implant tip in an Allen CCF subregion from 2D histology - Batch-register coronal histology with DeepSlice, detect each fiber tip, and read its Allen CCF subregion plus a hit/miss-vs-target verdict — a scripted QUINT replacement.
• Organize a raw DICOM dataset into a BIDS layout - Drive HeuDiConv or dcm2bids from Claude Code via the K-Dense BIDS skill to convert a raw DICOM dump into a validated BIDS dataset ready for BIDS-Apps and OpenNeuro submission.
• Parse FCS flow-cytometry files for downstream immunophenotyping - Use the FlowIO Claude skill to parse FCS 2.0/3.0/3.1 files into tidy DataFrames with channel categorisation and batch metadata extraction, ready for downstream immunophenotyping or QC.
• Plan a synthetic route for a target molecule - Drive the CovaSyn MCP retrosynthesis suite from Claude Code to propose buyable-starting-material routes to a target SMILES and capture them as a reproducible route table.
• Predict an antibody–antigen complex to map an epitope - Use the Boltz plugin to predict an antibody/nanobody–antigen complex and read off a candidate epitope — with explicit handling of the field's low success rate and unreliable confidence ranking.
• Profile shotgun metagenome taxa with Kraken2 and Bracken - Use the Kraken2 skill to classify shotgun metagenomic reads to taxa and re-estimate abundances with Bracken, with database and confidence-threshold choices recorded.
• Predict gene-knockout phenotypes with flux balance analysis - Use the COBRApy Claude Skill to load a genome-scale metabolic model and screen single/double gene deletions for growth phenotypes and essential genes.
• Predict RNA secondary structure and target-site accessibility - Use the ViennaRNA skill to fold an RNA, get its MFE structure, base-pair probabilities, and the accessibility of a candidate target site — for siRNA/sgRNA/ASO design or riboswitch analysis.
• Prepare the correct protonation state of a ligand before docking - Use the Rowan skill in Claude Code to predict macro-pKa and the dominant protonation state of ligands at physiological pH before docking or MD, as a reproducible prepared-ligand table.
• Prioritize targets within a disease via Open Targets - Drive the Open Targets MCP plugin to rank candidate targets within a disease across the four prioritisation pillars (precedence, tractability, doability, safety) and emit a cited target shortlist.
• Predict checkpoint-blockade response for a tumor from its biomarker profile - Use the ToolUniverse Immunotherapy Response Prediction skill to integrate TMB, MSI, PD-L1, HLA, and immune expression into an evidence-graded ICI response score.
• Prioritize tumor neoantigens for a personalized cancer vaccine - Use the Neoantigen Prediction skill to turn a patient's somatic variants, HLA type, and tumor RNA into a filtered, tiered neoantigen shortlist for vaccine design.
• Profile a cancer cohort's genomics with cBioPortal - Use the cBioPortal Claude Skill to query TCGA-scale studies for mutation frequency, oncoprints, TMB, and survival by genomic subgroup in plain language.
• Profile ChIP-seq or ATAC-seq signal around genomic features - Use the deepTools Claude Skill to turn aligned ChIP-seq/ATAC-seq BAMs into normalized bigWig tracks and TSS/peak-centered profile and heatmap figures.
• Profile a compound's polypharmacology from ChEMBL bioactivity data - Use the ChEMBL connector in Claude Code to pull every measured activity for one compound, group by target and assay, and surface the off-target profile a medicinal chemist needs before committing to a chemotype.
• Rank a compound library against a target by predicted binding affinity - Use the Boltz plugin's hosted small-molecule screen to rank a SMILES library against a target by predicted binding affinity — no local GPU — and treat the score as a binder/non-binder enrichment filter, not a quantitative ranker.
• Rank the conformers and tautomers of a small molecule with semi-empirical QM - Drive the ChemGraph MCP server to optimize and free-energy-rank enumerated tautomers and conformers at GFN2-xTB, emitting a populations table instead of one guessed structure.
• Reconstruct B-cell clonal lineages from AIRR-seq - Use the Immcantation BCR Analysis skill to cluster AIRR-seq reads into clonal families, quantify somatic hypermutation and selection, and infer germline-rooted lineage trees.
• Detect and rank druggable pockets on a protein structure - Enumerate surface cavities on an apo structure with fpocket, rank them for ligandability with P2Rank, and emit a docking box — one Claude Skill.
• Detect somatic copy-number variants from tumor sequencing - Use the CNVkit Claude Skill to call somatic gene amplifications and deletions from tumor WES/targeted-panel BAMs and export segmented, plotted copy-number profiles.
• Quantify bulk RNA-seq FASTQ into a gene-level counts matrix - Chain the fastp and Salmon skills, then aggregate with tximport, to turn raw bulk RNA-seq FASTQ into the gene-level counts matrix a DE analysis needs.
• Quantify a western blot from a scanned image - Use the Western Blot Quantification skill to detect bands, normalize target to a loading control, aggregate replicates, and emit a committed densitometry table.
• Register longitudinal medical scans to a common frame - Drive SimpleITK from Claude Code to rigidly then deformably align a patient's baseline and follow-up CT/MRI, propagate contours, and emit a reproducible transform + warped volume.
• Run bulk RNA-seq differential expression from a counts matrix - Use the PyDESeq2 Claude Skill to fit a negative-binomial GLM from a counts matrix and sample-condition table, returning ranked log2 fold-change and BH-adjusted p-values.
• Diagnose a rare disease from patient phenotypes - Use the ToolUniverse Rare Disease Diagnosis skill to turn a patient's HPO phenotypes and candidate variants into a tiered, ACMG-aligned differential with prioritized genes.
• Run first-pass QC on a single-cell RNA-seq dataset - Use Anthropic's single-cell-rna-qc skill to MAD-filter a 10x .h5 or AnnData .h5ad file and emit standard QC figures before downstream Scanpy or scvi-tools work.
• Run functional enrichment on a gene list - Use the gget Claude skill to run a gene list through Enrichr against GO, KEGG, Reactome, and disease libraries, then ask Claude to summarise the enriched terms back to the biology with explicit citations.
• Run a GWAS on case-control genotype data - Drive PLINK2 from QC through PCA-adjusted logistic-regression association on a case-control cohort, returning Manhattan-ready summary stats.
• Scan a therapeutic antibody for glycosylation sites - Use the Glycoengineering skill to scan an antibody sequence for N-glycosylation sequons and O-glycosylation hotspots before committing to a cell-line or developability campaign.
• Scan a protein for candidate CD4 T-cell (helper) epitopes - Use the MHC Class II Prediction skill to rank a protein's peptides for HLA-DR/DQ/DP binding, nominating CD4 T-helper epitope candidates for vaccine design.
• Scan a protein for candidate CD8 T-cell epitopes - Use the MHC Binding Prediction skill to rank a protein's peptides for MHC class I binding and presentation, nominating CD8 T-cell epitope and neoantigen candidates.
• Segment and quantify cells in a microscopy image - Use the Cellpose skill to segment cells or nuclei in fluorescence/brightfield images and extract per-cell counts, areas, and intensities to a committed table.
• Segment an organ or tumor in a medical image with nnU-Net - Drive nnU-Net's self-configuring framework from Claude Code to train and run a voxel-accurate segmentation model for an organ or lesion in CT/MRI, with 5-fold cross-validated Dice.
• Scan approved drugs for repurposing candidates against a disease - Given a disease, use Open Targets to rank associated targets, ChEMBL and PubChem for approved-drug bioactivity, and DrugBank for mechanism / indication / interaction context to produce a ranked shortlist of repurposing candidates.
• Screen a polypharmacy medication list for drug-drug interactions - Ground Claude in the DDInter skill to screen a multi-drug medication list for pairwise interactions with severity and mechanism, instead of asking the model from memory.
• Set up a protein molecular dynamics simulation in GROMACS from a PDB ID - Drive the GROMACS Copilot component of molecule-mcp to take a PDB ID through topology, solvation, ion neutralisation, minimisation, NVT/NPT equilibration, and a short production MD, with RMSD/RMSF/Rg analysis on the resulting trajectory.
• Source purchasable compounds for a hit list - Drive the ToolUniverse Chemical Sourcing skill to turn a shortlist of SMILES into a per-listing procurement table with purity, lead time, and analog fallbacks.
• Sort spikes from a Neuropixels recording end-to-end - Use the K-Dense neuropixels-analysis skill in Claude Code to take a raw SpikeGLX or Open Ephys Neuropixels recording through preprocessing, drift correction, and Kilosort4 to curated single units.
• Predict a protein–protein complex to map the binding interface - Use the Boltz plugin to co-fold two proteins into a complex and read off the candidate binding interface — sampling widely and ranking by consensus rather than trusting the top-1 confidence.
• Superpose two protein structures and quantify where they differ - Superpose a predicted model onto an experimental structure (or apo onto holo) with the PyMOL skill, then report global RMSD, per-residue deviation, and a rendered overlay.
• Test which immune populations differ between groups in a cytometry cohort - Chain three bioSkills cytometry skills to compensate, cluster, and test a spectral-flow or CyTOF cohort for differential abundance and state without pseudoreplication.
• Test whether a proposed RNA structure is actually conserved - Use the Covariation Analysis skill to test a proposed RNA secondary structure with R-scape, and to establish whether the alignment even has the power to reject it.
• Tile and stain-normalize a whole-slide image for ML - Use the histolab skill to turn a gigapixel H&E slide into a tissue-masked, stain-normalized tile dataset with a committed, re-runnable script.
• Track animal pose in behavioral video - Use the DeepLabCut skill in Claude Code to go from behavioral video to a filtered, quality-controlled keypoint table with kinematics and provenance recorded.
• Train a QSAR model from your own assay data and predict untested compounds - Use the ChemLint MCP in Claude Code to clean an assay CSV, featurize, train and cross-validate a QSAR model with an applicability domain, then score an untested library.
• Triage a compound's toxicology and hazard profile before committing to it - Drive the ToolUniverse Chemical Safety skill to build a graded, source-attributed toxicology dossier per compound, distinguishing no-data from no-hazard.
• Triage a stack of new preprints in your field - Use the bio-research plugin in Claude Code to pull last-week bioRxiv and PubMed hits, rank by relevance, and produce a one-paragraph readout per paper.
• Trace an adverse outcome pathway from a molecular initiating event to an organ finding - Drive the ToolUniverse Adverse Outcome Pathway skill to assemble candidate MIE-to-endpoint pathways for an observed toxicity, with weight of evidence recorded per link.
• Triage an AlphaFold model for structure-based drug design - Pull a UniProt AlphaFold prediction, surface pLDDT regions and pocket-residue confidence, and produce a go/no-go verdict on whether the model is fit for downstream docking or modelling — all from one MCP server.
• Fit a survival model to censored clinical outcomes - Use the scikit-survival Claude Skill to fit Cox and Random Survival Forest models to censored time-to-event data and report concordance with proper validation.
• Predict the regulatory effect of a non-coding variant - Use the DeepMind AlphaGenome skill to score how a single non-coding variant changes expression, chromatin accessibility, histone marks, splicing, and TF binding.
• Scan adverse-event reports for a drug-safety signal - Use the OpenFDA MCP server to pull FAERS adverse-event reports and label warnings for a drug, rank the top reported reactions, and assemble a pharmacovigilance snapshot.
• Score a drug-combination screen for synergy - Drive the ToolUniverse Drug Synergy skill to pick the right reference model (Bliss / HSA / Loewe / ZIP / Chou-Talalay) for your combination data and classify a two-drug pairing as synergistic, additive, or antagonistic.
• Score point mutations for functional impact with a protein language model - Use the ESM skill to compute zero-shot masked-marginal log-likelihood-ratio scores for a list of point mutations, ranking them by predicted functional effect without any labelled data.
• Validate a drug target with a GO/NO-GO score before committing bench work - Drive the ToolUniverse Drug Target Validation skill to score one hypothesized target 0–100 across genetics, druggability, safety, and clinical precedent, emitting a cited GO/NO-GO card.
• Vet a PDB structure before you build on it - Judge whether a deposited structure — or one specific loop or side chain in it — is reliable enough to dock against, measure, or build a mechanism on.
• Get balanced reactions and cross-references for a metabolite or EC number - Have Claude Code script the Rhea release files so a metabolite name or EC number returns mass- and charge-balanced reactions with direction and cross-references.
• Handle missing endpoint data in a confirmatory trial - Use the Missing Data Sensitivity skill to fix an estimand, run MMRM as primary, and build a reference-based and tipping-point sensitivity package.
• Define a phenotype cohort in an OMOP CDM database - Build a frozen OMOP concept set with OMOPHub, then count and characterise the cohort in your CDM with pyomop, emitting a portable attrition table.
• Predict hospital readmission from EHR data - Use the PyHealth Claude Skill to build a 30-day readmission predictor on MIMIC/OMOP EHR data with a reproducible dataset → task → model → metrics pipeline.
• Predict ADMET properties for a compound series with an ML predictor - Drive the ADMETlab MCP server to batch-predict ~119 ML ADMET endpoints (with uncertainty) for a SMILES series, then commit a triaged CSV, pinned client env, and a provenance record.
• Screen a bacterial genome for resistance and virulence genes - Use the BLAST MCP server to search a bacterial assembly's proteins against CARD and VFDB, emitting a curated resistance- and virulence-gene profile with a committed script and pinned databases.
• Track single particles and measure their diffusion coefficient - Use the trackpy skill to locate and link particles across video-microscopy frames, then compute MSD and diffusion coefficients into a committed trajectory table.
• Triage GWAS lead SNPs to candidate drug targets - Drive GWAS-MCP to annotate a list of GWAS lead SNPs — variant consequence, eQTL-implicated genes, GWAS Catalog trait context, and drug-target/tractability support — into a cited variant-to-target triage table.
• Crucible evaluation protocol (T1–T6)
• Generate falsification-tested drug-repurposing hypotheses across corpora - Use the Hypothesis Crucible plugin to mine typed knowledge fragments from literature, Open Targets/ChEMBL/ClinicalTrials, and raw experimental data (GEO), bridge them Swanson-style into candidate drug-disease connections, then aggressively reject the non-novel, ungrounded, contradicted, and implausible — surfacing only survivors with a fragment-cited mechanism and a falsifiable experiment.
• Reproducible example: Hypothesis Crucible
• Reproducible example: functional enrichment
• Reproducible example: glycosylation-altering variants
• Reproducible example: implant-tip localization