Claude Skills by HolobiomicsLab

Use when after feature detection has produced a feature table with zero

Use when running an LC-MS or GC-MS preprocessing and statistics pipeline

Use when publishing a metabolomics dataset into the MetabolomicsHub index

Use when assembling a reanalysis or meta-analysis cohort from public

Use when moving a feature table between Workflow4Metabolomics tools and

Use when an agent needs to find and apply a computational-metabolomics / LC-MS-MS skill from this collection, and optionally ground it against the source paper via Perspicacité before acting.

Use when an ASB skill proved wrong, stale, missing or wasteful in practice — its steps failed, no skill covered the task, the leaves existed but nothing composed them, or the tool has changed. Turns that friction into a redacted, dedupable report the user approves before anything is filed.

Use when starting any task with the ASB Metabolomics skill collection — read this meta-skill first. It explains good practice (search -> apply -> ground), enforces the license-tier acknowledgment for non-open tools, then hands off to the _router skill for actual skill selection.

Use when a user has a whole metabolomics analysis GOAL (e.g. "annotate my untargeted LC-MS/MS data", "find biomarkers", "where else has this molecule been seen") rather than a single step — select the right end-to-end composite workflow super-skill, then run its stages, grounding each against its source papers.

'Use when you want chemical-class-level annotations for untargeted LC-MS/MS

'Use when you have untargeted LC-MS/MS MS2 data and want to spread a

'Use when you have GC-MS data (mzML / CDF, typically EI) and want deconvolved,

'Use when you have a genome-scale constraint-based metabolic model (GEM,

'Use when you want to train or apply a graph neural network over molecular

'Use when you have a parent structure (drug, natural product, xenobiotic)

'Use when you have ion-mobility LC-IMS-MS/MS data (e.g. timsTOF / PASEF)

'Use when you have untargeted lipidomics LC-MS/MS data (mzML) and want

'Use when you have a spectrum or feature of interest and want to know

'Use when you have mass-spectrometry imaging data (imzML, e.g. MALDI/DESI)

'Use when you want to discover shared substructures (Mass2Motifs) across

'Use when you have NMR metabolomics data (1D/2D spectra or FIDs) and

'Use when you have paired genomic and metabolomic data from the same

'Use when you have an LC-MS metabolomics feature list (m/z, optionally

'Use when you have MS/MS for unknown features (a SIRIUS-flavour mgf /

'Use when you want to annotate untargeted MS2 spectra with a machine-learned

'Use when you have LC-MS data from a stable-isotope (e.g. 13C / 15N)

'Use when you have a metabolomics feature/quant table and want a statistically

'Use when you have untargeted HRMS data and want to screen for a defined

'Use when you have targeted LC-MS data for a defined panel of analytes

'Use when you have untargeted LC-MS/MS data (mzML) and want an annotated

Use when after running DESeq() and extracting raw results with results(), when you have log fold change estimates with high variance and wish to improve their precision.

Use when immediately after calling squidpy.gr.spatial_neighbors() or similar spatial graph construction methods on an AnnData object. It is essential when validating that the computed spatial graph has been correctly stored in adata.

Use when you have PSI (percent-spliced-in) matrices calculated independently for two or more biological conditions, each with two or more replicate samples, and you want to identify which alternative splicing events show statistically significant changes in inclusion levels between conditions.

Use when when you have a GTF genome annotation and need to identify all local alternative splicing events (SE, RI, A5/A3, MX, AF/AL) or transcript-level isoform events for a given gene set, prior to quantifying PSI values across samples or performing differential splicing analysis.

Use when after executing a Squidpy spatial analysis function (e.g., gr.spatial_neighbors, gr.nhood_enrichment, gr.sepal, im.

Use when you have single-cell RNA-seq count matrices or processed expression data and need to store them alongside cluster assignments (e.g., leiden cluster labels), cell metadata, and computed analysis results (e.

Use when after applying Scanpy preprocessing functions (e.g., pp.normalize_total, pp.pca) to a Dask-backed AnnData object, or when performing any operation that could alter matrix dimensions, data types, or backing storage (dense, sparse, or lazy).

Use when after running DESeq2 differential expression analysis and extracting results with raw log fold changes, apply this skill when you observe high variance in effect size estimates across genes—particularly when many genes have small counts, unreliable variance estimates, or when you want.

Use when you have transcript-level quantification files (quant.gz, h5, or similar) from a known upstream quantifier (salmon, kallisto, sailfish, oarfish) and need to import them into R as matrices for differential expression analysis with edgeR, DESeq2, or limma-voom.

Use when you have raw FASTQ sequencing reads (single-end or paired-end) and a reference transcriptome FASTA file, and you need to determine which transcript(s) each read aligns to in order to quantify transcript abundance. This is the core mapping stage in a salmon quant workflow;

Use when when two mapping implementations (or versions of the same mapper) show disagreement on per-read mapping status—e.g., one mapper leaves reads fully unmapped that the other maps, or one maps with high confidence where the other is uncertain.

Use when when quantifying or mapping RNA-seq reads with salmon quant using the --writeMappings (-z) flag, or in any streaming output scenario where record count discrepancies appear between reported totals (e.g., NumReads in quant.sf) and file contents (SAM record count).

Use when when a C++ program writes records to an output stream (e.g., SAM alignment file) and the final output file contains fewer records than expected based on upstream counts (e.g., salmon's NumReads total in quant.sf exceeds SAM record count), indicating buffered data loss at stream closure.

Use when you have (1) spatial omics data loaded in AnnData format with a pre-built spatial neighbor graph (from squidpy.gr.spatial_neighbors() or similar), (2) a categorical variable in the AnnData object (e.g., cell type, tissue compartment, annotation stored as .obs or .

Use when when comparing mapped read counts between two RNA-seq quantification implementations (e.

Use when you have cloned a scientific Python project (e.g., scverse/scanpy) and need to verify that your local development environment matches the CI specification before submitting contributions, or when auditing whether the published test suite executes without failures on a fresh checkout.

Use when after fitting a linear model to expression data using limma's lmFit function on a design matrix encoding experimental groups, inspect the resulting MArrayLM object to retrieve coefficient estimates (log-fold-changes) and standard errors needed to assess which genes show meaningful.

Use when you have raw read count matrices from RNA-seq quantification (e.g., from featureCounts, HTSeq, Salmon, or kallisto) and need to prepare them for differential expression analysis.

Use when after constructing a count matrix from transcript quantification files (via tximport, HTSeq, featureCounts, or direct alignment) and before running DESeq() differential expression analysis.

Use when when applying Scanpy preprocessing functions (e.g., pp.normalize_total, pp.pca) to AnnData objects where the expression matrix X is backed by a dask.array.Array, you need to verify that the operation completed without eagerly loading the full matrix.