> For the complete documentation index, see [llms.txt](https://docs.hits.ai/hyperlab-catalog/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.hits.ai/hyperlab-catalog/skill-catalog-en/all-en/hyperlab-all-skills-catalog.md). # HyperLab All Skills Catalog Explore all HyperLab skills, including their functions, input requirements, and supported scope.

Category	Skill	Description
Genomics and Bioinformatics	anndata-data-structure	Holds single-cell data in the standard container, the expression matrix with per-cell and per-gene annotation, alternative layers, embeddings and neighbour graphs, and reads and writes it to disk. It organises; the analysis runs on top.
Genomics and Bioinformatics	arboreto-grn-inference	Infers which transcription factors likely regulate which genes from expression data, scoring every factor-target link by importance rather than a p-value and writing the network out. Scales to single-cell sized matrices, though very large ones need trimming to the most variable genes first.
Genomics and Bioinformatics	archs4-database	Looks up expression for a human or mouse gene across more than a million public RNA sequencing samples, all processed the same way, so tissue patterns and co-expressed genes compare directly without reprocessing anything yourself.
Genomics and Bioinformatics	bakta-genome-annotation	Annotates a bacterial or archaeal genome: coding sequences, the various RNA genes, small open reading frames, CRISPR arrays and origins of replication, written out in the formats a public archive accepts. The first run needs several minutes to download its reference database, and only the smaller version fits here, leaving more genes with generic descriptions.
Genomics and Bioinformatics	bcftools-variant-manipulation	Filters, merges, normalises and annotates variant files after calling, quality thresholds, multi-sample merges, adding reference identifiers, pulling genotypes out, and computes summary statistics.
Genomics and Bioinformatics	bedtools-genomic-intervals	Does arithmetic on genomic intervals: finding overlaps, merging regions, computing coverage, locating the nearest feature, and pulling the underlying sequence out, across the common coordinate formats.
Genomics and Bioinformatics	biopython-molecular-biology	Handles everyday sequence work: reading and converting the common formats, manipulating sequences, querying public archives, running searches, and working with structures. The general toolkit for batch jobs and custom pipelines.
Genomics and Bioinformatics	biopython-sequence-analysis	Compares sequences to each other, pairwise and multiple alignment, then building distance-based phylogenetic trees (neighbour-joining or UPGMA) and working with them, with retrieval from the public archives wired in. For gene family and cross-species work.
Genomics and Bioinformatics	bioservices-multi-database	Reaches dozens of biological web services through one programmatic interface, so a script can move between proteins, pathways, compounds and identifiers without learning each service separately. Built for cross-database workflows.
Genomics and Bioinformatics	busco-status-interpretation	Explains what a completeness score for a genome, transcriptome or predicted protein set is actually telling you: why duplicated genes still count as complete, how to compare assemblies fairly, and the counting mistakes that inflate or deflate the number.
Genomics and Bioinformatics	bwa-mem2-dna-aligner	Aligns DNA sequencing reads to a reference genome and writes sorted alignments ready for variant calling, handling reads that span a breakpoint. Suited to bacterial and other small genomes; a human genome index is too large to build here. For RNA, a splice-aware aligner is needed instead.
Genomics and Bioinformatics	cbioportal-database	Pulls cancer genomics from large patient cohorts, somatic mutations, copy number, expression, and the clinical record including survival, across several hundred published studies.
Genomics and Bioinformatics	celltypist-cell-annotation	Labels cell types from pretrained reference models covering immune, gut, lung, brain, fetal and tumour tissue, giving per-cell labels, a majority label per cluster, and a confidence score. Fast, and one model at a time.
Genomics and Bioinformatics	cellxgene-census	Searches a public collection of tens of millions of annotated cells by cell type, tissue, disease or organism, and streams what you select without loading it all. For a reference dataset you did not generate yourself.
Genomics and Bioinformatics	clinpgx-database	Looks up how a patient's genotype should change their prescription: the curated dosing guidelines, gene-drug pairs, variant-drug associations, and the regulatory label text standing behind them.
Genomics and Bioinformatics	clinvar-database	Looks up what a variant is clinically known to mean, whether it is considered pathogenic, which conditions it is tied to, who submitted the assessment and how thoroughly it was reviewed.
Genomics and Bioinformatics	cnvkit-copy-number	Detects copy number gains and losses from aligned targeted or whole-genome sequencing reads, normalising coverage against matched or pooled normal samples, segmenting it, and calling amplifications and deletions with estimates of tumour purity. The first run needs a few minutes of setup, including a reference genome download.
Genomics and Bioinformatics	dbsnp-database	Looks up a known variant by its reference identifier, gene or region, returning its alleles, what kind of change it is, how common the minor allele is, and links out to the other variant resources.
Genomics and Bioinformatics	deeptools-ngs-analysis	Turns alignments into normalised coverage tracks and the plots that go with them, signal profiles and heatmaps around features, sample correlation and principal components, for ChIP, ATAC and RNA sequencing.
Genomics and Bioinformatics	depmap-crispr-essentiality	Reads DepMap CRISPR knockout screens across cancer cell lines to say where a gene is essential and where it is not, correlating its expression with how essential it is while handling the missing values properly.
Genomics and Bioinformatics	deseq2-differential-expression	Finds the genes that change between your conditions from raw bulk RNA sequencing counts, shrunken fold changes, corrected p-values, and the volcano, MA and heatmap views that go with them. Multi-factor designs included. This is the R route.
Genomics and Bioinformatics	ena-database	Searches and downloads sequence data from the European archive, raw reads, assemblies and annotated sequences, returning the file locations so a public dataset can be pulled down directly.
Genomics and Bioinformatics	encode-database	Searches regulatory genomics experiments across a thousand cell types, transcription factor binding, open chromatin, histone marks, expression, and downloads the peak and signal files, or the catalogued regulatory elements over a region.
Genomics and Bioinformatics	ensembl-database	Looks up gene, transcript and variant annotation in Ensembl across more than three hundred species: a gene by symbol or identifier, its sequence, its orthologs in other species, its regulatory features, cross-references to other naming systems, and the predicted effect of a variant.
Genomics and Bioinformatics	etetoolkit	Works with phylogenetic trees once you have them: reading the standard formats, walking and annotating the nodes, attaching taxonomy, and rendering a figure you can put in a paper.
Genomics and Bioinformatics	fastp-fastq-preprocessing	Checks raw Illumina sequencing reads and cleans them in one pass, detecting adapters, trimming them off, dropping low-quality reads, correcting overlapping pairs, and writes a quality report. The first step before any alignment.
Genomics and Bioinformatics	featurecounts-rna-counting	Counts how many reads fall on each gene, taking aligned reads and a gene annotation and producing the per-gene matrix that differential expression needs. Handles strandedness and paired reads across many samples at once.
Genomics and Bioinformatics	gatk-variant-calling	Calls germline single-nucleotide variants and short insertions and deletions from duplicate-marked aligned sequencing data, following the best-practice route: per-sample calling, joint genotyping across the cohort, then recalibration or hard filters. Suited to a few samples rather than a large cohort; the first run takes extra time to download the matching reference and known-variant files.
Genomics and Bioinformatics	gene-database	Resolves a gene to its official record: the approved symbol, its aliases, reference sequence identifiers, a summary of what it does, and where it sits, across a very wide range of species.
Genomics and Bioinformatics	geniml	Turns genomic regions into vectors a model can use, training and applying region embeddings, indexing interval datasets and searching that space. For clustering samples by accessibility or classifying regulatory elements. The first run needs a few minutes of setup.
Genomics and Bioinformatics	geo-database	Searches the public expression repository and pulls a study down for reanalysis, the series matrix, the platform annotation and the per-sample metadata, loaded as a table you can work with. For RNA-seq studies the counts often sit in supplementary files rather than the series matrix.
Genomics and Bioinformatics	gget-genomic-databases	Answers a question about one gene from many resources with a single command: its record and sequence, known structures, enrichment, disease and drug links, cancer and single-cell data. Quick lookups rather than deep queries.
Genomics and Bioinformatics	gnomad-database	Looks up how common a variant is in each major ancestry group, and how much a gene tolerates being broken, so a variant can be judged rare and a gene judged constrained.
Genomics and Bioinformatics	gseapy-gene-enrichment	Tests whether a gene list is concentrated in known pathways or signatures, either by over-representation on a cut list or by running the score down a fully ranked one, and returns the enrichment tables and plots.
Genomics and Bioinformatics	gtars	Processes very large collections of interval files quickly, reading and writing them, intersecting, merging, complementing and subtracting, and reduces regions to a shared vocabulary for later modelling.
Genomics and Bioinformatics	gwas-database	Looks up which traits and diseases a variant has been associated with in published genome-wide studies, or which variants are reported for a disease, with the effect sizes and the studies behind them, down to the summary statistics where a study has deposited them.
Genomics and Bioinformatics	harmony-batch-correction	Removes batch effects from single-cell data after dimensionality reduction and before clustering, while guarding against over-correction that would erase real biological differences. Linear and fast, scaling to millions of cells.
Genomics and Bioinformatics	homer-motif-analysis	Finds which transcription factor motifs are over-represented in a set of peaks, both known motifs and ones discovered from the sequence itself, and reports what each peak sits near. The step after peaks are called. The first run needs a few minutes of setup to download the genome sequence.
Genomics and Bioinformatics	jaspar-database	Looks up the binding motif for a transcription factor as a position matrix, by name, species or structural family, and scans a sequence for where that motif occurs. The motif itself, not observed binding.
Genomics and Bioinformatics	macs3-peak-calling	Finds the regions where sequencing signal piles up against background, whether transcription factor sites, histone marks or open chromatin, and writes them out as peaks: narrow for sharp binding, broad for diffuse marks.
Genomics and Bioinformatics	monarch-database	Connects diseases, genes and phenotypes into one graph you can walk, including across species, so a set of observed phenotypes turns into a list of candidate genes ranked by phenotype overlap.
Genomics and Bioinformatics	mouse-phenome-database	Looks up measured phenotypes for inbred mouse strains, strain-level means or individual animals, across hundreds of projects, searchable by phenotype term, so a strain can be chosen on evidence.
Genomics and Bioinformatics	multiqc-qc-reports	Gathers the quality metrics that over 150 standard pipeline tools leave behind and merges them into one interactive report, so dozens of samples can be compared on one screen instead of opening each log.
Genomics and Bioinformatics	plink2-gwas-analysis	Runs a genome-wide association study over large genotype datasets by linear or logistic regression, with the quality control, relatedness estimation and population structure correction done first, and association statistics ready to plot.
Genomics and Bioinformatics	popv-cell-annotation	Labels cell types by running about ten different methods against a labelled reference and taking the majority vote, reporting each method's call and how far they agreed. For when one method's answer is not enough to trust. The first run needs a few minutes of setup, and the deep-learning methods train slowly here, so large datasets may need to skip them.
Genomics and Bioinformatics	prokka-genome-annotation	Annotates a prokaryotic genome, bacteria, archaea and viruses, finding coding sequences and RNA genes against the established reference families. The long-standing route that older pipelines expect. The first run needs a few minutes of setup.
Genomics and Bioinformatics	pydeseq2-differential-expression	Runs bulk RNA differential expression from a raw count table and sample sheet inside a Python pipeline with no R involved: normalisation, negative binomial fitting, corrected p-values, shrunken fold changes, and the usual plots. Two-group and multi-factor comparisons.
Genomics and Bioinformatics	pysam-genomic-files	Opens alignment, variant and sequence files from inside a script, so you can pull the reads over one region, walk a pileup, or apply a filter the standard tools do not offer.
Genomics and Bioinformatics	quickgo-database	Looks up Gene Ontology terms and the proteins annotated to them, a term by identifier or keyword, its place in the ontology above and below, and annotations filtered by species or evidence. The step before an enrichment test.
Genomics and Bioinformatics	regulomedb-database	Scores how likely a human variant is to affect gene regulation, from strongest evidence down to none, and shows what it overlaps: factor binding, histone marks, open chromatin, footprints, motifs and expression effects.
Genomics and Bioinformatics	remap-database	Looks up where transcription factors have actually been observed binding, from a large collection of published experiments, the peaks over a region or near a gene, filtered by factor, species or element type.
Genomics and Bioinformatics	roary-pangenome	Compares annotated bacterial genomes to work out which genes they all share and which belong to only some, producing the presence-and-absence matrix, a core-gene alignment, and a pan-genome reference. The first run needs a few minutes of setup.
Genomics and Bioinformatics	salmon-rna-quantification	Quantifies transcript and gene expression without aligning to the genome, matching reads against a transcriptome index in minutes and correcting for sequence and GC bias. Much faster, but leaves you no alignments to inspect.
Genomics and Bioinformatics	samtools-bam-processing	Sorts, indexes, filters and summarises alignment files, the step between aligning reads and calling anything from them. Marks duplicates, merges runs, and reports depth and alignment statistics.
Genomics and Bioinformatics	scanpy-scrna-seq	Takes single-cell RNA data through the standard route: quality control, normalisation, variable gene selection, dimensionality reduction, neighbour graph, clustering and UMAP, then finds the markers defining each cluster.
Genomics and Bioinformatics	scikit-bio	Measures diversity in microbial communities from an OTU or ASV count table, within and between samples, with and without a phylogenetic tree, and tests whether groups genuinely differ, with the ordination plots that go alongside. Also builds phylogenetic trees from distance matrices.
Genomics and Bioinformatics	scvi-tools-single-cell	Models single-cell data with deep generative networks: batch correction, semi-supervised annotation, joint RNA and surface protein, and transfer onto a reference, with uncertainty attached. Heavier than a linear correction, and more flexible. Best suited to datasets under about 50,000 cells, and the first run needs a few minutes of setup.
Genomics and Bioinformatics	single-cell-annotation-guide	Decides how to label clusters before you start: naming them from canonical markers, transferring labels from a reference, or doing both and checking one against the other. Covers tissue matching, doublets, and clusters that resist naming.
Genomics and Bioinformatics	snpeff-variant-annotation	Predicts each variant's functional effect, how severe it is, the gene and transcript affected, the amino acid change, then filters the list down and adds known clinical significance. The first run needs a few minutes of setup to download the genome annotation database.
Genomics and Bioinformatics	star-rna-seq-aligner	Aligns RNA sequencing reads across splice junctions, finding novel junctions as well, with an optional second pass for more sensitivity, and returns sorted alignments, the junction table and, if requested, per-gene counts. The route when you need the alignments themselves. Suited to smaller genomes; a human or mouse genome index is too large to build here.
Genomics and Bioinformatics	ucsc-genome-browser	Pulls sequence, gene models and annotation tracks for any region across a hundred-odd assemblies, plus cross-species conservation scores where the assembly has them. The route when you want what the browser shows, as data.
Genomics and Bioinformatics	vcf-variant-filtering	Checks whether a variant file has been filtered yet and applies quality thresholds before any summary is computed, so counts and transition-to-transversion ratios are not read off raw calls. Also says when not to filter.
Structural Biology and Drug Discovery	admet	Predicts the ADME/T profile of a small molecule from its SMILES: solubility, permeability, brain penetration, P450 and hERG liability, microsomal stability, basic pKa, LogD and drug-likeness. It profiles the molecule; it says nothing about affinity for a target.
Structural Biology and Drug Discovery	aizynthfinder-retrosynthesis	Works out how to make a molecule, searching backwards from the target through known reaction types until it reaches building blocks you can buy, and ranks the routes it finds by length and how many starting materials are in stock. Reaction conditions and yields are not predicted, and finding no route does not mean the compound cannot be made.
Structural Biology and Drug Discovery	alphafold-database-access	Fetches any of the two hundred million protein structures AlphaFold has already predicted, by UniProt entry, with the per-residue and pairwise confidence scores that say which parts to trust. Predictions, not experiments.
Structural Biology and Drug Discovery	chembl-database-bioactivity	Pulls measured bioactivity out of ChEMBL, IC50, Ki and EC50 values for compounds against targets, so you can find what inhibits a target, build a series, or look up a drug's mechanism and indications.
Structural Biology and Drug Discovery	clinicaltrials-database-search	Searches registered clinical trials by condition, drug, phase, sponsor or location, and pulls the full record for any trial by its registry number. For tracking who is testing what, and how far along.
Structural Biology and Drug Discovery	dailymed-database	Looks up the official FDA label for a drug: what it is approved to treat, how it is dosed, its warnings and adverse reactions, and how it is packaged. This is the label text itself, not the post-market reports.
Structural Biology and Drug Discovery	datamol-cheminformatics	Cleans and compares chemical structures with sensible defaults already set: standardising them, computing descriptors and fingerprints, clustering, picking a diverse subset, breaking them into scaffolds and fragments.
Structural Biology and Drug Discovery	ddinter-database	Looks up whether two drugs interact, how serious it is, by what mechanism, and what the clinical advice is. Covers a few hundred thousand interaction pairs across approved drugs.
Structural Biology and Drug Discovery	emdb-database	Looks up cryo-EM density maps along with the atomic models fitted into them, resolution, method, organism and sample for each entry, and where to download the map. For coordinates alone, a different resource applies.
Structural Biology and Drug Discovery	fda-database	Searches what the FDA has recorded after a drug reached the market, adverse event reports (largely from 2004 onward), recalls and enforcement actions, searchable by drug, ingredient or reaction term, and can also pull the drug's label. Report counts reflect voluntary reporting, not incidence.
Structural Biology and Drug Discovery	gtopdb-database	Looks up how a known ligand acts on a receptor, its binding affinity, whether it agonises or antagonises, and the species it was measured in, from the pharmacology community's curated reference.
Structural Biology and Drug Discovery	mdanalysis-trajectory	Analyses a finished molecular dynamics trajectory from any of the common simulation packages: RMSD, per-residue fluctuation, radius of gyration, contact maps and hydrogen bonds, with flexible selection of the atoms you care about.
Structural Biology and Drug Discovery	mdtraj-trajectory-analysis	Computes the standard measures from a molecular dynamics trajectory, RMSD over time, radius of gyration, per-residue fluctuation, residue contact maps, backbone torsions for Ramachandran plots, and eight-state secondary structure.
Structural Biology and Drug Discovery	medchem	Filters a compound list down to what is worth pursuing, drug-likeness rules such as Lipinski, and structural alerts such as PAINS that flag frequent hitters and awkward chemistry, with the filters composable to your own criteria.
Structural Biology and Drug Discovery	molfeat-molecular-featurization	Turns a molecule into the numeric features a model can learn from, fingerprints, computed descriptors, pretrained chemical embeddings, pharmacophores, behind one interface instead of a dozen. The first run needs a few minutes of setup to install its machine-learning libraries, and pretrained embeddings also download their model weights.
Structural Biology and Drug Discovery	opentargets-database	Reports how well a target is linked to a disease through the Open Targets score, which pulls evidence from many sources into a single number, alongside the drugs already aimed at that target and what is known about its safety and tractability.
Structural Biology and Drug Discovery	pdb-database	Searches experimentally determined protein structures and downloads them, by text, by sequence or by 3D similarity, with the metadata that tells you how good each one is.
Structural Biology and Drug Discovery	pignet	Scores how tightly a ligand binds in a pose you already have, in kcal/mol, broken down into physical energy terms. It ranks poses; it does not generate them.
Structural Biology and Drug Discovery	pubchem-compound-search	Looks up a compound by name, identifier, structure or formula and returns its properties, weight, LogP, polar surface area, hydrogen-bond counts, and runs similarity or substructure searches across the collection.
Structural Biology and Drug Discovery	pytdc-therapeutics-data-commons	Supplies curated, model-ready datasets for drug discovery, absorption and toxicity, drug-target and drug-drug interaction, with splits and metrics already standardised so benchmark numbers compare.
Structural Biology and Drug Discovery	rdkit-chemdraw-cdxml	Reads and writes ChemDraw files, including the parts a structure toolkit normally loses, reaction arrows, plus signs, scheme steps and text labels, so a synthesis scheme survives the round trip. Each file comes with a rendered PNG preview, since opening it in ChemDraw itself is not tested.
Structural Biology and Drug Discovery	rdkit-cheminformatics	Works with chemical structures in detail, parsing them, computing descriptors and fingerprints, ranking compounds by similarity to a reference, clustering them for diversity, searching substructures, judging drug-likeness, enumerating reactions, generating coordinates. The fine-grained route when defaults are not enough.
Structural Biology and Drug Discovery	sar-analysis	Relates what a compound looks like to how active it is: it finds the shared scaffold of your analogue series, splits each compound into R-groups and shows which substituents at each position raise or lower potency. Needs a table of SMILES and potency for four or more analogues, and returns an R-group table with a potency heatmap.
Structural Biology and Drug Discovery	unichem-database	Translates a compound identifier between the major chemical databases, resolving a structure key to whichever source identifiers exist for it, and finding compounds that share the same connectivity.
Structural Biology and Drug Discovery	zinc-database	Looks up purchasable and make-on-demand compounds in the ZINC library, a substance by its identifier, a structure resolved to its entry, which catalogues carry it, and assembles a set for docking.
Protein Design and Structure Prediction	antibody-design	Runs an antibody or nanobody campaign end to end, CDR backbone generation, sequence design, then structure prediction, and reports the candidates that pass, ranked by binding confidence and structural self-consistency. Requires an antibody or nanobody framework structure to graft the CDRs onto.
Protein Design and Structure Prediction	kfold	Predicts the 3D structure of a complex, protein with a ligand, DNA, RNA or another protein, and, for protein-ligand complexes only, estimates binding affinity alongside it for ranking ligands against one target. Also folds a designed sequence to check it takes the shape you intended.
Protein Design and Structure Prediction	kfold-refine	Cleans up a predicted structure so the local geometry is physically sensible, checking a ligand's chemistry against its SMILES and relieving atom clashes and strained ligand geometry without pulling it away from the prediction. Worth running before docking or rescoring, though it fixes geometry rather than a wrong pose.
Protein Design and Structure Prediction	ligandmpnn	Designs sequence around a bound ligand, metal or nucleic acid, choosing residues that fit what they pack against. For enzyme active sites, metal sites and protein-DNA/RNA interfaces; full all-atom side-chain models are optional.
Protein Design and Structure Prediction	peptide-design	Designs peptides, linear or head-to-tail cyclic, as target binders or standalone, and validates them structurally, including the extra handling cyclic peptides need before they can be folded at all. Natural amino acids only.
Protein Design and Structure Prediction	pocket-redesign	Redesigns only the residues lining a binding pocket, keeping the backbone fixed and the ligand in place throughout, then checks the redesigned pocket still holds it. For enzyme active sites and shifting substrate specificity.
Protein Design and Structure Prediction	protein-design	Runs a de novo design campaign end to end, backbone generation, sequence design, then structure prediction to see which designs hold up, and reports the candidates that survived filtering. For binders, monomers, scaffolds and oligomers; antibodies are handled by the antibody design workflow.
Protein Design and Structure Prediction	proteinmpnn	Designs amino acid sequences that fold stably into a given backbone, with control over which chains and positions get redesigned and how much sequence diversity to sample.
Protein Design and Structure Prediction	rfantibody-diffusion	Generates new CDR loop backbones for an antibody or nanobody against the epitope you specify, on the framework structure you provide. The backbones carry placeholder sequences, sequence design is the next step.
Protein Design and Structure Prediction	rfantibody-mpnn	Designs the CDR sequences on antibody backbones, leaving the framework untouched. Works on backbones from a CDR backbone generation run or on your own antibody structures with CDR residues annotated.
Protein Design and Structure Prediction	rfdiffusion	Generates new protein backbones for a design goal, a binder to your target, a monomer, a motif scaffold, a symmetric oligomer or a cyclic peptide. Shape only; the sequence comes at the next step.
Proteomics and Metabolomics	interpro-database	Looks up which domains a protein contains and which family it belongs to, drawing several classification resources into one answer, with the taxonomic spread of each family. A way in when the function is unknown, for proteins already in UniProt rather than a new sequence.
Proteomics and Metabolomics	matchms-spectral-matching	Identifies a metabolite by matching its MS/MS spectrum against a reference library you supply, importing the common spectrum formats, cleaning the peaks and scoring similarity several ways. For putting a name to an unknown peak.
Proteomics and Metabolomics	metabolomics-workbench-database	Searches the metabolomics studies NIH publishes and the reference metabolite records behind them, finding a study, resolving a compound to its standard name, or searching by precursor mass.
Proteomics and Metabolomics	pride-database	Searches the PRIDE proteomics archive and downloads from it: projects by organism, instrument, disease or software, their metadata, their raw and processed files, and which projects mention a given protein.
Proteomics and Metabolomics	pyopenms-mass-spectrometry	Processes raw LC-MS/MS data for proteomics or untargeted metabolomics, reading mzML or mzXML files, smoothing and picking peaks, detecting and linking features across samples, then filtering peptide hits from a separate search engine at a controlled error rate and inferring proteins. Vendor raw files must be converted to mzML first.
Proteomics and Metabolomics	uniprot-protein-database	Looks up a protein in UniProt by gene or protein name, its sequence, what is known about its function, and the identifiers that connect it to the other resources you are using.
Systems Biology and Multiomics	cellchat-cell-communication	Infers which cell types are signalling to which from single-cell expression with cell-type labels, ranking the ligand-receptor pairs (human or mouse) between each pair of cell groups. The full CellChat package cannot be set up here, so pathway-level summaries and sender and receiver role scores are not produced.
Systems Biology and Multiomics	cobrapy-metabolic-modeling	Predicts metabolic flux through a genome-scale model, what the network can produce, which genes are essential, what a knockout does, from the constraints alone, without needing kinetic rates. Needs an existing model of the organism (your own or one from a public collection such as BiGG); results depend on the chosen objective and medium.
Systems Biology and Multiomics	kegg-pathway-analysis	Guides KEGG pathway enrichment on a differential expression result: choosing between over-representation and rank-based testing, splitting up and down regulation rather than pooling them, and reading the output honestly.
Systems Biology and Multiomics	lamindb-data-management	Tracks where each dataset came from and what was done to it, versioning the artifacts, recording lineage across steps, and validating against standard ontologies, so each result in an analysis can be traced back to its source data. The record stays within that analysis and is not carried over to other analyses.
Systems Biology and Multiomics	libsbml-network-modeling	Builds, edits and validates biological network models in the standard exchange format, reactions and their kinetics, species, rules, and the extension used for flux balance, so a model moves between tools intact. Also updates existing published models, such as those from BioModels, and converts between format versions.
Systems Biology and Multiomics	mofaplus-multi-omics	Finds the shared axes of variation across several omics layers measured on the same samples, decomposing them into factors, then reports how much variance each factor explains in each layer and what the factors track with.
Systems Biology and Multiomics	muon-multiomics-singlecell	Analyses two or more measurements taken from the same cells together, RNA with chromatin accessibility, or RNA with surface protein, keeping each modality's processing separate before joining them into one embedding.
Systems Biology and Multiomics	omics-analysis-guide	Decides which level of omics analysis a question actually needs: a validated pipeline, a standard workflow, or a method built for the case. Covers already-quantified bulk RNA-seq and proteomics data, with guidance on sample QC, missing-value handling, test choice and false-discovery control.
Systems Biology and Multiomics	reactome-database	Looks up curated biological pathways and tests whether a gene list is enriched in any of them, with the pathway hierarchy and cross-references to other naming systems.
Systems Biology and Multiomics	string-database-ppi	Retrieves the known and predicted interactions around a protein, builds the network, and tests whether the set interacts more than chance would give. Scores combine experiments, curated databases and text mining.
Cell Biology and Imaging	cellpose-cell-segmentation	Outlines individual cells and nuclei in microscopy images using pretrained models that carry across stains and modalities without retraining. Returns label masks you can measure or track. The first run needs a few minutes of setup, and large image sets or 3D stacks are slow to process.
Cell Biology and Imaging	flowio-flow-cytometry	Reads and writes flow cytometry data files, handing you the events as arrays with their channel metadata and exporting to a table. This handles the file itself; gating and compensation are a separate step.
Cell Biology and Imaging	opencv-bioimage-analysis	Processes bio-images quickly, colour conversion, morphology, contour and blob detection, template matching and optical flow, at the speed large batches and time-lapse video need.
Cell Biology and Imaging	pyimagej-fiji-bridge	Drives ImageJ and Fiji from a script so existing macros and plugins run unattended and their results come back as arrays. The route when a lab protocol is already written as a Fiji macro. The first run needs a few minutes to download Fiji, and macros that wait for on-screen clicks or dialogs cannot run.
Cell Biology and Imaging	scikit-image-processing	Filters, segments and measures microscopy images, thresholding, watershed, active contours, and region properties such as area and intensity. The general-purpose route for quantifying what is in an image.
Cell Biology and Imaging	trackpy-particle-tracking	Tracks particles through a video frame by frame and joins their positions into trajectories, fluorescent spots, vesicles, whole cells, then computes mean squared displacement to give diffusion coefficients. Works in 2D and 3D, and needs the pixel size and frame rate to report diffusion in physical units.
Molecular Biology	plannotate-plasmid-annotation	Finds the features in a plasmid sequence, promoters, terminators, resistance genes, origins and tags, and returns an annotated GenBank file with an interactive map. Reads circular sequences correctly, so whole vectors work, not just linear inserts. The first run needs a few minutes of setup to install the tool and download its feature database.
Molecular Biology	sgrna-design-guide	Picks sgRNAs for a CRISPR experiment, preferring guides already validated in the literature, then precomputed genome-wide designs, and designing from scratch only when the target has neither. Guides designed from scratch follow sequence rules only, without genome-wide off-target screening.
Molecular Biology	viennarna-structure-prediction	Predicts how an RNA sequence folds on itself, the minimum-free-energy secondary structure and how confident each base pair is, and how two RNAs pair with each other. Covers siRNA and sgRNA targeting, and checking whether a site is accessible.
Medical Imaging	histolab-wsi-processing	Prepares a whole-slide pathology image for analysis: finds the tissue, cuts it into tiles, and discards the blank and unusable ones. Turns a slide archive into a training set, or screens a batch of scans first. The first run needs a few minutes of setup.
Medical Imaging	imaging-data-commons	Searches the cancer imaging collections the NCI publishes and downloads what you select, radiology and pathology studies, browsable by collection, patient, study and series, with no account needed. Whole collections can run to hundreds of gigabytes or more, too big to download in full here, so select the series you need.
Medical Imaging	pydicom-medical-imaging	Reads and writes DICOM images from CT, MRI, X-ray and ultrasound, pixels as arrays, tags you can edit, windowing, and stacking a series into a 3D volume. Removes the main patient-identifying header fields, but study dates, private tags and text burned into the image remain, so this is not full de-identification.
Medical Imaging	simpleitk-image-registration	Aligns 3D medical images taken at different times or on different machines, and segments regions within them, from rigid through deformable registration, with conversion between the common medical formats.
Biostatistics	pymc-bayesian-modeling	Fits a Bayesian model so the answer arrives with its uncertainty, priors you set, posterior sampling, convergence diagnostics and model comparison. Suits small samples, grouped or nested data through hierarchical models, or analyses where prior knowledge should count.
Biostatistics	scikit-survival-analysis	Models time to an event when some subjects never reach it, using Cox models, survival forests and boosting on censored data, and scores how well the model ranks risk. Covers Kaplan-Meier and competing risks.
Biostatistics	statistical-analysis	Decides which statistical test your data calls for, checks the assumptions it rests on, and reports the result with effect size and power in APA style. Covers frequentist tests and Bayesian alternatives.
Biostatistics	statsmodels-statistical-modeling	Fits classical statistical models and tells you whether the terms matter, ordinary and generalised regression, logistic and count models, time series, with the diagnostics and hypothesis tests that belong with them.
Scientific Computing and Data	aeon	Classifies, clusters and segments time series, sensor traces, waveforms, anything ordered in time, predicts continuous outcomes or extracts features from them, and copes with panels, multiple channels, and series of unequal length.
Scientific Computing and Data	astropy-astronomy	Handles astronomical data: units that carry their dimensions, conversions between celestial coordinate systems, observation files and tables, cosmological distances, and precise astronomical time scales.
Scientific Computing and Data	dask-parallel-computing	Spreads a computation too big for memory across the available cores in chunks, with parallel versions of the dataframe and array you already use. It runs on a single machine here, not across a cluster. Reach for it when the data no longer fits in memory.
Scientific Computing and Data	degenerate-input-filtering	Strips out the inputs that would break or quietly corrupt a statistical test, constant features, zero-variance columns, all-missing columns, features detected in too few samples, empty files, single-sequence alignments, before the test runs, and reports how many were removed and why.
Scientific Computing and Data	exploratory-data-analysis	Opens a data file you have not seen before and works out what it is, how good it is, and what you could reasonably do with it, across tables, sequences, images, spectra and omics.
Scientific Computing and Data	geopandas-geospatial	Analyses geographic data and draws it on a map: spatial joins and overlays, distances and areas, conversion between coordinate systems, and choropleth or interactive maps.
Scientific Computing and Data	nan-safe-correlation	Correlates thousands of features that carry missing values without the matrix shortcuts that silently give wrong answers, deleting pairwise where it should and staying fast on large inputs. Each result comes with its count of usable samples, and features with fewer than 10 are skipped.
Scientific Computing and Data	neb-irc-activation-energy	Finds the transition state of a reaction and reports its activation barrier, searching the minimum-energy path, optimising the saddle point, then confirming it with an intrinsic reaction coordinate. Returns an energy profile and a mode animation. Barriers are approximate, from a fast semi-empirical method suited to ranking related reactions and practical for systems up to about 150 atoms.
Scientific Computing and Data	networkx-graph-analysis	Analyses a network: which nodes are hubs by several measures of centrality, the shortest paths through it, and which nodes group into communities. Suits graphs up to tens of thousands of nodes.
Scientific Computing and Data	neurokit2	Processes physiological recordings and extracts the standard measures, heart rate and its variability, brain-wave complexity and power, muscle activation onset, skin conductance responses.
Scientific Computing and Data	nextflow-workflow-engine	Builds a multi-step pipeline, or adapts an nf-core community one, that runs the same way on a laptop, a compute cluster or the cloud, with every step inside its own container. Here steps run directly rather than in containers; cluster and cloud runs happen on your own infrastructure. This is the route when you need portability across environments.
Scientific Computing and Data	polars-dataframes	Handles a large table fast while it still fits in memory, filtering, grouping, joining and pivoting tables of several gigabytes where older tools would crawl. Data that exceeds memory needs a different approach.
Scientific Computing and Data	pyhealth	Builds prediction models from electronic health records, mortality, readmission, diagnosis, drug recommendation, handling MIMIC and other standard clinical datasets that you supply, and the medical coding systems they use. The first run needs several minutes of setup.
Scientific Computing and Data	pymatgen	Analyses crystal structures and materials properties: reading and building structures, pulling computed properties from the Materials Project, phase and Pourbaix diagrams, and simulated diffraction patterns. Materials Project lookups, which the phase and Pourbaix diagrams draw on, need your own API key.
Scientific Computing and Data	pymoo	Optimises several competing objectives at once, raising activity while lowering toxicity, say, and returns the Pareto front, so you see the whole trade-off rather than one forced compromise. You supply the functions or models that score each objective; it does not predict activity or toxicity itself.
Scientific Computing and Data	scikit-learn-machine-learning	Trains and evaluates the standard machine-learning models, classification, regression, clustering, dimensionality reduction, with preprocessing and tuning wired into a single pipeline.
Scientific Computing and Data	shap-model-explainability	Explains why a model predicted what it did by attributing the result to individual features, and ranks which features matter overall. Works on tree, deep, linear and black-box models, though deep-learning models need a few minutes of extra setup on first use.
Scientific Computing and Data	simpy-discrete-event-simulation	Simulates a process where work queues and resources are limited, a lab's throughput, a sample pipeline, a service desk, so you can find the bottleneck before changing anything for real.
Scientific Computing and Data	snakemake-workflow-engine	Chains analysis steps together by their input and output files into a reproducible pipeline that runs the same steps on every sample, working out the order and what can run in parallel, and rerunning only what changed. This is the route when your pipeline is rule-based.
Scientific Computing and Data	sympy-symbolic-math	Does algebra and calculus exactly rather than numerically, derivatives, integrals, limits, equation solving, matrices and differential equations, and turns the result into runnable code.
Scientific Computing and Data	umap-learn	Compresses data with thousands of features down to two or three dimensions you can plot, keeping both local neighbourhoods and the broader shape. Supervised and density-aware variants are available.
Scientific Computing and Data	uspto-database	Searches United States patent data by inventor, assignee, classification or keyword, and pulls the metadata, abstracts and claim counts so you can map a portfolio or check prior art.
Scientific Computing and Data	vaex-dataframes	Works through a table too large to load comfortably into memory by reading it off disk in pieces, aggregating, filtering and plotting its rows without ever loading them whole. Suited to files of up to about 10 GB rather than terabytes, and the first run needs a few minutes of setup. Tables that do fit in memory have a faster option.
Scientific Computing and Data	zarr-python	Stores a large multi-dimensional array in compressed chunks so you can read just the slice you need, from local disk or cloud storage, without touching the rest of it.
Data Visualization	matplotlib-scientific-plotting	Draws a publication-quality figure with control over every element, line, scatter, bar, heatmap, contour and 3D, in multi-panel layouts, exported as PNG, PDF or SVG. Slower to write than the shortcuts, but nothing is out of reach.
Data Visualization	molecular-visualization-3dmol	Turns a molecular structure into a standalone web page you can rotate in a browser, sticks, spheres, cartoon or surface, and animates trajectories or vibrational modes with play and speed controls. One file, no viewer to install.
Data Visualization	multipanel	Assembles several plots into one journal-ready figure with panels A, B and C, either redrawing every panel into a single unified layout or compositing images you already rendered onto one canvas. Exports vector PDF and high-resolution PNG.
Data Visualization	omics-plotting	Supplies the house style and ready-made recipes for omics figures, volcano, MA, heatmap, GSEA, box and violin, PCA and UMAP scatter, Kaplan-Meier, Manhattan and forest, so every figure in an analysis looks like one system.
Data Visualization	plotly-interactive-plots	Builds a chart you can hover, zoom and pan in a browser, a volcano plot that names the gene under the cursor, a dose-response you can explore, an expression heatmap. Exports as a web page or as a static image.
Data Visualization	scientific-visualization	Helps you decide how to show a result before anything is plotted, which chart fits the data, which colours survive printing and colour blindness, how to compose the panel, and what the common mistakes are.
Data Visualization	seaborn-statistical-plots	Draws a statistical summary in a line or two, distributions, box and violin plots, regression fits, correlation heatmaps, faceted grids, with the aggregation and confidence intervals computed for you.
Data Visualization	statistical-significance-annotation	Adds significance markers to a comparison plot the way journals expect, from ns through four asterisks, with brackets placed over the right pairs and the thresholds stated in the legend.
Research Writing	cancer-research-figure-guide	Checks your figures against Cancer Research requirements from AACR, 300 to 1200 DPI, EPS, TIFF or AI, hierarchical panel labels such as Ai and Aii, and limits on how many figures and tables you may include. Legends must state replicate counts.
Research Writing	cell-figure-guide	Checks your figures against Cell Press requirements, 300 to 1000 DPI, TIFF or PDF, RGB colour, Avenir or Arial type, uppercase panel labels, under one of the strictest image-manipulation policies in the field.
Research Writing	citation-management	Compares the main reference managers and shows how to apply the citation style a journal asks for, whether APA, Vancouver, ACS or Nature, and keep DOIs consistent across Word, Google Docs and LaTeX.
Research Writing	clinical-decision-support-documents	Writes clinical decision support documents: biomarker-stratified cohort analyses and treatment reports carrying GRADE evidence grades from 1A to 2C, with hazard ratios, confidence intervals and survival reported the way reviewers expect. Cohort analyses need trial or cohort data to work from, and every document covers patient groups rather than individual treatment plans.
Research Writing	elife-figure-guide	Checks your figures against eLife requirements, TIFF, EPS or PDF, a striking image at 1800 by 900 pixels, and the naming scheme for figure supplements. Selective enhancement counts as misconduct, not a formatting slip.
Research Writing	general-figure-guide	Reviews a finished plot for the faults that get figures sent back, overlapping labels, clipped text, missing axes or legends, overcrowded data, and checks resolution and format against what journals generally expect.
Research Writing	hypothesis-generation	Turns an observation into three to five competing hypotheses you can actually test, the mechanism each proposes, the predictions that tell them apart, and the experiments that would decide between them. This weighs several rival ideas against each other rather than sharpening just one.
Research Writing	lancet-figure-guide	Checks your figures against The Lancet's requirements, 300 DPI at 120 percent, editable PowerPoint, Word or SVG, column widths of 75 or 154 mm, Times New Roman. The Lancet redraws figures in house.
Research Writing	latex-research-posters	Builds a conference poster in LaTeX from one of the standard poster templates, covering layout, typography, colour, figure placement and accessibility. It compiles the poster to PDF and keeps the LaTeX source, so it stays editable.
Research Writing	literature-review	Plans and runs a literature review, a PRISMA protocol, a documented search in PubMed with matching search strings for you to run in Scopus, Embase and Web of Science, screening and extraction, then synthesis as narrative, thematic or meta-analysis. A publishable systematic review still needs at least two independent human screeners.
Research Writing	nature-figure-guide	Checks your figures against Nature's requirements, at least 300 DPI, AI, EPS or TIFF, RGB colour, Helvetica or Arial type, lowercase panel labels, and explains which image adjustments their image-integrity policy allows.
Research Writing	nejm-figure-guide	Checks your figures against NEJM requirements, 300 to 1200 DPI and editable vector art in AI, EPS or SVG, and reminds you to remove patient-identifying details. NEJM creates medical illustrations with its own illustrators, so send a sketch or description rather than finished artwork.
Research Writing	peer-review-methodology	Reviews a manuscript or grant the way a referee would, in seven passes, first impression, section by section, statistics, reproducibility, figure integrity, ethics, and the writing, and returns a structured report.
Research Writing	pnas-figure-guide	Checks your figures against PNAS requirements, 300 to 1000 PPI, TIFF, EPS or PDF, RGB only, Arial or Helvetica type, italicised uppercase panel labels. It also explains what PNAS's automated image screening looks for.
Research Writing	science-figure-guide	Checks your figures against Science's requirements, 150 to 300 DPI at initial submission and at least 300 DPI for the revised manuscript, PDF, EPS or TIFF, RGB colour, Myriad or Helvetica type. Their manipulation policy is strict: even a gamma adjustment has to be declared.
Research Writing	scientific-brainstorming	Generates research ideas by structured method rather than free association, drawing on SCAMPER, Six Thinking Hats, morphological analysis, TRIZ and biomimicry, and picks the method to suit whether you are stuck or refining.
Research Writing	scientific-critical-thinking	Weighs what a scientific claim is actually worth, where the study sits in the evidence hierarchy, what the effect size means next to the p-value, what might be confounding it, and which biases the design leaves open.
Research Writing	scientific-manuscript-writing	Drafts a manuscript in IMRAD structure, in the citation style your target journal uses, and against the reporting checklist your study type requires, whether CONSORT, STROBE, PRISMA or ARRIVE.
Research Writing	scientific-schematics	Plans an illustrative figure before anyone draws it, which form fits, whether mechanism, pathway, workflow or graphical abstract, which tool to build it in, and how colour and layout should carry the meaning.
Research Writing	scientific-slides	Shapes a talk for a conference, seminar, thesis defence or grant pitch, what belongs on each slide, how the argument is ordered, how long each part runs, and how to make data readable from the back of the room.
Deliverables	hyperlab-generate-figure	Draws a schematic, pathway or graphical abstract as an editable vector figure, so you can still move arrows and fix labels in Illustrator or PowerPoint afterwards. Cells, organelles and other illustrated subjects are AI-generated images placed inside it rather than editable shapes. For diagrams, not data plots.
Deliverables	hyperlab-html-infographic	Condenses your results into a one-page visual summary, figures, key numbers and takeaways designed as a single page you open in a browser. One self-contained file that works offline.
Deliverables	hyperlab-hwpx	Reads, summarises, edits, fills in and creates Hancom Hangul documents (.hwpx), keeping the document's own formatting. Older .hwp files are read with a format-aware fallback.
Deliverables	hyperlab-methods	Drafts the Methods section for the computational analysis you ran in this conversation, with parameters, versions and each published tool cited to a real PMID or DOI. Anything the record could not confirm is listed as a gap rather than filled in.
Deliverables	hyperlab-poster	Lays your results out as a single-page conference poster, sized to the format you need, delivered as an editable PowerPoint file.
Deliverables	hyperlab-pptx	Builds an editable PowerPoint presentation from your results, with slides laid out and figures placed for you, and can redo individual slides later. Upload a .pptx deck of your own with real slides (not a .potx or master-only template) and it will follow that instead.
Deliverables	hyperlab-report	Turns the analysis from this conversation into a written report you can hand to someone, a summary opening with an AI-generated graphical abstract of the key findings, then background, methods, results, figures and limitations, delivered as a PDF. Every number traces back to what was actually run.
Laboratory Automation	protocolsio-integration	Searches protocols.io for published wet-lab, bioinformatics and clinical protocols by keyword or identifier and pulls back the whole procedure: steps, reagents, materials, equipment and timing.
Laboratory Automation	western-blot-quantification	Quantifies the bands on a western blot, detecting them, normalising the signal, and applying the statistics that blot data calls for, so the numbers hold up when someone asks.
Skills	hyperlab-create-skill	Helps you write a new skill of your own, interviewing you over a few rounds about what it should do and checking first that no existing skill or built-in capability already covers it. The skill is saved to your account, and existing ones are extended rather than duplicated.
Skills	hyperlab-find-skills	Searches the open skill ecosystem (skills.sh, GitHub, npm) for an existing skill that does what you need, inspects it, and adds it to your account if it fits. When an outside skill will not run as is, it hands off to adapting it or writing a new one.
Skills	hyperlab-save-as-skill	Turns an analysis you have just finished in this conversation into a reusable skill, generalising the specific files and values into parameters so you can run the same workflow on new data. If a matching skill already exists, it offers to update that one instead.