Claude Skills by HolobiomicsLab

Use when you have a merged and batch-corrected non-targeted LC-MS/MS feature table with sample metadata and need to perform statistical testing to identify significant features, compare groups (e.g., disease vs. control), or explore multivariate patterns.

Use when after training a customized R statistical or machine learning model on annotated training data, you need to persist the trained model object for reuse in downstream analysis workflows without retraining.

Use when you have raw mass spectrometry data in mzXML, mzML, or netCDF format from untargeted LC/HRMS analysis that has been pre-processed by IDSL.

Use when you have raw CE-MS or LC-MS instrument files (stored as OnDiskMSnExp objects or similar Bioconductor containers) and need to extract quantitative features (migration times, m/z values, peak intensities) by orchestrating multiple R packages in a controlled, documented sequence.

Use when when applying a series of mpactr filter functions (filter_mispicked_ions, filter_group, filter_cv, filter_insource_ions) with copy_object=FALSE to confirm that the original peak table object is mutated as intended, not silently copied.

Use when when working with large metabolomics peak tables (e.

Use when you need to understand or validate whether calling filter_mispicked_ions() (or similar R6 filter methods) with different copy_object settings will mutate your original data object in memory or preserve it.

Use when your MetaboSet object contains missing values (marked as NA) in the expression matrix after quality flagging, but you need complete data for multivariate analysis.

Use when when you have predictions from one or more metabolite annotation models (MLP baseline, GNN, or ESP ensemble) and need to quantify ranking performance on ESI/LC-MS test spectra.

Use when you have completed cluster-based filtering of KEGG candidate assignments in untargeted LC-MS metabolomics and need to rank those candidates by biological plausibility using a metabolite interaction network. Specifically, use it after `clusterBased.

Use when you have a feature quantification table (m/z, retention time, peak areas) and candidate metabolite annotations (chemical identifiers, MS/MS spectra matches, or SIRIUS/GNPS predictions) linked to a sample organism or taxon, and you need to prioritize which candidates are most biochemically.

Use when after running annotateRC on LC–MS AIF data when you need to inspect whether a feature has multiple plausible annotations (e.g., isobaric lipids, isomers with the same fatty-acyl chain) or when the rank-1 annotation confidence is borderline and alternatives should be evaluated.

Use when after generating ranked predictions of chemical formulas or subformulas for MS/MS spectra (from a neural network model like MIST-CF's formula transformer), compare predicted assignments against ground-truth reference labels to measure ranking quality.

Use when after running retention-order prediction experiments on a test or held-out evaluation dataset.

Use when you have multiple pre-trained neural network models (e.g., MLP and GNN) that produce overlapping predictions on the same set of candidates, and your evaluation metric is rank-based (average rank, Rank@K) rather than point-wise accuracy or RMSE.

Use when you have raw GC-MS output files (vendor formats or netCDF) from a chromatography instrument and need to prepare them for automated peak deconvolution and spectral analysis.

Use when you have a Thermo Fisher Scientific .raw file (e.g., Q Exactive HF, Orbitrap) and need to extract specific spectral scans, chromatographic traces, scan-level metadata, or file-level headers programmatically—e.

Use when you have native Thermo Fisher RAW files and need to recover file-level metadata (instrument details, run statistics), scan headers (retention time, total ion current, scan mode for MS1 or MS2), or peak lists (m/z and intensity arrays) in tabular or array form suitable for computational.

Use when beginning an LC-MS data analysis pipeline and you need to rapidly inspect instrument metadata, acquisition parameters, or scan statistics from proprietary Thermo .raw files without the I/O overhead of loading full spectra.

Use when you have raw UPLC-HRMS data from ThermoFisher or Agilent instruments and need to feed it into MSThunder for nontargeted pollutant identification. Your input is a vendor binary format (.raw or .d) that MSThunder cannot directly ingest. Environment constraints (e.

Use when you have Thermo Orbitrap .raw files and need to access raw spectral data (individual MS1 or MS2 scans, base-peak values, chromatogram traces, retention times, or scan-level metadata) for custom analysis, visualization, or integration into an R-based pipeline.

Use when after feature detection and alignment (XCMS or equivalent), when you have a CSV feature table with m/z and retention time columns and need to group features derived from the same compound.

Use when you have molecular structure data (SMILES strings or MOL files) from an in-house chemical database and need to prepare it as input for a graph neural network that predicts liquid chromatography retention times for small molecule identification.

Use when you have loaded intracellular metabolomics data (LC-MS normalized abundances) and a stoichiometric metabolic model with multiple reactions, and you plan to compute reaction propensity scores or other flux-related metrics that depend on substrate concentrations.

Use when you have computed RAS (Reaction Activity Scores) from transcriptomics and GPR rules, RPS (Reaction Propensity Scores) from intracellular metabolomics via mass-action kinetics, and flux distribution differences (FFD) from constraint-based sampling across multiple biological samples.

Use when after CREST (version >= 3.0.2) has identified an ensemble of low-energy conformers and stationary points (minima and transition states), and before submitting interpolated geometries to ORCA (version >= 6.0.

Use when you have measured intracellular metabolite abundances (LC-MS or similar) across multiple cell lines or conditions and a stoichiometric metabolic model (with reaction-metabolite associations) to estimate how differences in substrate availability—independent of gene expression—translate into.

Use when a QC check fails during an active LC-MS instrument run and you need to immediately notify configured users (via Slack channel or email address) of the failure event, including timestamp, check type, and severity metadata.

Use when you have a trained NeatMS neural network model and labelled peak validation data, and need to select an operational classification threshold or understand how TPR and FPR vary across probability thresholds (e.g., 0.00–0.99).

Use when when you have computed similarity scores (cosine, modified cosine, Spec2Vec, or other metrics) between a set of query spectra and a reference library with known structural annotations (InChIKey), and you need to evaluate how well each scoring method distinguishes true structural matches.

Use when after alignment of two or more same-method nontargeted LCMS datasets (e.g., via Eclipse) when you have a feature table containing retention times and feature intensities across samples. Use it when redundancy is expected—e.

Use when you have detected feature tables from multiple LC-IMS-MS/MS samples (each with mz, drift_time, retention_time, and intensity columns) and need to match corresponding features across datasets to enable cross-sample quantitation or cohort analysis.

Use when you have executed a structure annotation pipeline (like BAM) on a validation dataset for which ground-truth molecular structure annotations exist, and you need to assess whether the pipeline's predictions match the reference annotations at the required sensitivity and specificity.

Use when when applying sequential filters to a large metabolomics peak table (e.g., mispicked ions, group, CV, or in-source filters) and you need to confirm that setting copy_object=FALSE actually modifies the input object in-place rather than creating a hidden copy.

Use when when you have raw count tables from 16S rRNA sequencing (microbiome) or LC-MS/MS metabolomics (metabolome) and need to train neural network or regression models for microbe-metabolite relationship prediction.

Use when apply this filter when working with MS-DIAL peak lists (feature tables with m/z, retention time, and intensity) that contain features with anomalous mass defects—particularly when you have prior knowledge of the expected RMD range for your sample type or analytical method, or when you want.

Use when you have a peak table from XCMS preprocessing with intensity measurements for the same set of metabolites across multiple QC replicate injections (samples marked SampleType='LQC'), and you need to filter out EICs with poor reproducibility before evaluating peak quality or training a.

Use when after frequency-based denoising has been applied to individual replicate spectra within each feature (via generate_denoised_spectra), you have a collection of denoised fragment ion lists per feature per scan.

Use when you have multiple MS/MS spectra (replicates) for a single metabolic feature (same m/z and RT window) and need to identify which fragments are reproducibly detected across replicates versus noise.

Use when after extracting raw MS/MS spectra from mzML files for individual features (identified by precursor m/z and retention time) and you have multiple replicate spectra for the same feature that need to be pooled for consensus analysis.

Use when when you have a USI string (e.g., 'mzspec:GNPS:TASK-abc123:scan:1943') and need to retrieve the underlying spectrum data from its native repository without knowing a priori which backend stores it.

Use when when starting a fresh ENPKG installation, you have a GitHub URL (e.g., https://github.com/enpkg/enpkg_full or https://github.

Use when you have identified a published method (e.g., MIST-CF for chemical formula ranking from mass spectra) whose source code and trained weights are available in a public repository, and you need to apply that method to new experimental data (e.

Use when after running FIDDLE v2.0.0 inference on MS/MS spectra and obtaining ranked formula candidates with confidence scores, apply this skill when the rescore model outputs columns named Rescore (0), Rescore (1), ...

Use when when you have TCN-predicted candidate formulas with ranked scores and need to train a Siamese rescore model to re-rank those candidates.

Use when when you have an MS2 spectrum (FT2 scan or mzML) with observed peak data (m/z, intensity, charge states), a peptide sequence, and need to annotate which amino acid residues generated each matched fragment ion—particularly in stable isotope labeling experiments where isotopic mass shifts.

Use when when you have paired LC-MS data from both labeled (isotope-tracer dosed) and unlabeled (control) samples of the same analytes, and you want to correct FAM to true MDV while accounting for resolution effects from the mass spectrometer (especially relevant for high-resolution instruments.

Use when when you have a USI string (comprising dataset identifier, spectrum index, and optional library reference) and need to generate a stable, machine-readable link that resolves to interactive spectrum visualization or programmatic access.

Use when after combining multiple EI or MS2 mass spectral libraries and you have access to NIST RI reference files (ri.dat and USER.DBU) and need to assign experimental retention indices to compounds.

Use when you have a compiled EI or MS2 library object (from read_lib or c() combination of multiple sources) and a local NIST library installation with accessible ri.dat and USER.DBU files in the mssearch/nist_ri directory.