Claude Skills by HolobiomicsLab

Use when after discovering Mass2Motifs via LDA topic modeling on MS/MS data, when you need to assign putative structural annotations to those motifs by comparing their fragmentation signatures (weighted fragment and neutral loss distributions) to a curated reference database of known motifs and.

Use when you have a query spectrum (or set of query spectra) and need to rank candidate library spectra by their likelihood of sharing the same chemical structure.

Use when you have a collection of preprocessed and cleaned mass spectrometry spectra (in mzML, mzXML, msp, MGF, or JSON format) and need to compute all-pairs or targeted spectral similarity scores to identify related spectra, perform spectral library searches, or build a similarity network.

Use when after generating TP candidates (via in-silico prediction or library lookup) and extracting MS/MS peak lists for both parent features and TP feature candidates, use spectral similarity scoring to quantify fragmentation pattern overlap.

Use when when you have pre-trained embedding vectors (e.g., from MSBERT) for a collection of mass spectra and need to establish which spectra are chemically similar for validation, clustering, or library matching.

Use when when preparing MS/MS spectra from .msp files for transformer-based deep learning models in IDSL_MINT. Specifically: you have raw spectral data with variable peak counts and need fixed-size tensor inputs;

Use when when converting pre-processed MS/MS spectra into fixed-length vector representations using Word2Vec embeddings for Spec2Vec similarity scoring. Specifically, apply this skill after you have represented individual peaks and neutral losses as words ('[redacted-email]', 'loss@xxx.

Use when when you have a collection of MS/MS reference spectra and unknown query spectra that must be rapidly matched against a large spectral library, and you need to enable approximate nearest neighbor indexing to reduce computational cost from exhaustive pairwise comparison to K-nearest neighbor.

Use when you have loaded multidimensional MS data (from MZA HDF5 files or other formats) and need to examine a specific m/z region—for example, to visualize a known lipid or metabolite mass range, perform peak detection within a narrow window, or reduce computational overhead by working on a subset.

Use when after implementing or modifying a cross-language integration layer that wraps Python mass spectrometry functions (e.g., spectral matching, peak detection, normalization from matchms or spectrum_utils) as R-callable functions within Spectra workflows.

Use when you have tandem MS spectra from structurally related or known compounds and need to decide which similarity metric will maximize correct ranking of related molecules in a spectral library search.

Use when after generating probability predictions for potential modification sites (via ModiFinder.

Use when you have a TCN-predicted training set of MS/MS spectra with formula annotations and need to prepare it for Siamese rescore model training.

Use when you have raw, high-resolution MS/MS spectra in mzML, mzXML, or MGF format that need to be prepared for fast similarity searching or clustering.

Use when you have computed a sparse pairwise distance matrix from nearest neighbor indexes and need to group spectra into clusters. Use this skill when: (1) you have a sparse similarity or distance matrix as input;

Use when you have acquired raw or semi-processed mass spectra and need to search them against curated domain-specific databases (microbeMASST, plantMASST, etc.) using standalone web applications or batch APIs.

Use when when you have a USI (e.g., mzspec:MTBLS1124:QC07.mzML) pointing to a public mzML or related spectrum file in MetaboLights, MassIVE, or GNPS repositories, and need to load the spectrum data for interactive visualization, quality control assessment, or downstream analysis.

Use when after embedding MS/MS spectra into 32-dimensional GLEAMS vectors, when you need to group spectra by the same peptide origin.

Use when when implementing or auditing a deep learning pipeline for MS/MS-based molecular formula prediction, verify that precursor m/z values in the input feature array are zeroed before they reach the spectrum encoder (e.g., TCN).

Use when you have raw or minimally processed tandem MS spectra (in mzML, mgf, or other standard formats) and need to prepare them for spectral matching, library searching, or quantitative analysis.

Use when when you have raw mass spectrometry data from diverse instrument vendors (Thermo, Sciex, etc.) and need to harmonize and standardize spectrum-level metadata—including scan information, precursor m/z and charge, and ion selection parameters—into a queryable, vendor-agnostic tabular schema.

Use when when you have a USI (Universal Spectrum Identifier) string referencing a spectrum in a public metabolomics repository (GNPS Molecular Networking, GNPS Spectral Libraries, MassBank, MetaboLights, Metabolomics Workbench, MS2LDA, or ProteoXchange) and need to extract its raw spectral data.

Use when a user submits one or more MS/MS spectra (via .mgf file, USI list, or direct upload) and the downstream analysis requires dispatching to a specific domain-specific MASST tool (microbeMASST, plantMASST, tissueMASST, microbiomeMASST, foodMASST, or metadataMASST).

Use when you have raw MS2 spectra in common formats (mzML, mzXML, msp, MGF, JSON) from one or more metabolomics samples, and you need to prepare them for MS2 fingerprint generation, similarity scoring, or cross-sample comparison.

Use when when you have raw tandem mass spectrometry peak data (m/z and intensity pairs), precursor m/z, charge state, and adduct annotation for one or more compounds, and need to construct normalized spectrum objects suitable for downstream spectral alignment, modification site prediction, or.

Use when you have pairs of MS/MS spectra and need to estimate their structural similarity (Tanimoto score based on molecular fingerprints) as a proxy for compound relatedness.

Use when you have a test set of annotated MS/MS spectra with known structural similarity (via molecular fingerprints or InChIKey), and you want to assess how well a spectral similarity measure (learned or classical) retrieves structurally related compound pairs across a full range of thresholds.

Use when when you need to quantify and compare the filtering efficacy of mutually exclusive noise-threshold methods on the same input mass spectrum, or when validating that a selected noise-filtering strategy retains an expected number of peaks for downstream molecular formula assignment.

Use when you have an experimental tandem mass spectrum (peaks with m/z values and intensities), a chemical formula, and a set of candidate molecular structures retrieved from a database like PubChem.

Use when when you have a tandem mass spectrum (MS/MS) and a ProForma 2.0 peptidoform string (e.g., DLTDYLM[Oxidation]K) and need to identify which observed spectrum peaks correspond to expected b-ion and y-ion fragments, in order to validate peptide identification or annotate spectrum quality.

Use when you have raw MS/MS spectra in MGF or other standard formats that need to be ingested into a machine learning pipeline for cross-modal matching against molecular structures, or when spectra from different collision energy levels or instruments require standardization before comparative.

Use when when preparing raw MS/MS spectra for input to a Siamese neural network trained to predict structural similarity scores (Tanimoto).

Use when when you have raw MS/MS spectra with residual noise or low-intensity peaks and plan to calculate spectral entropy, entropy similarity, or perform spectral library matching.

Use when when training a formula-prediction model with a frozen pretrained TCN spectrum encoder and unfrozen FormulaEncoder and RescoreHead components, or when you suspect the model may use precursor intensity as a spurious feature rather than fragment-pattern information for molecular formula.

Use when you have raw mass spectrometry spectra (in MGF, mzML, or similar formats) that must undergo standardized preprocessing before library matching, similarity searching, or performance benchmarking.

Use when you have parsed LC-MS/MS spectral data (precursor m/z, ionization mode, collision energy, and a list of fragment m/z and intensity pairs) and need to submit it to the CSI:FingerID web service for molecular fingerprint prediction as part of a metabolite identification workflow.

Use when you have a collection of MS/MS spectra from reference standards representing your compounds of interest (e.g., flavonoids, prenylated chalcones) and a set of MS/MS spectra from non-target or other compounds.

Use when when you have thousands to millions of high-resolution tandem MS/MS spectra (in mzML, mzXML, or MGF format) that need to be clustered or compared, and exhaustive pairwise distance computation is computationally prohibitive.

Use when after duplicate filtering of MZmine-exported MGF and CSV files, when you have combined spectra from multiple samples in a single MGF and need to segregate them by sample identifier before fragment annotation or adduct assignment.

Use when after successfully constructing a nearest neighbor index from hashed spectrum feature vectors and before performing density-based clustering or similarity searches.

Use when when clustering large collections (thousands to millions) of tandem mass spectra and you need to compute a sparse pairwise distance matrix for density-based clustering.

Use when when you have a processed or annotated MsmsSpectrum object (from USI loading or direct instantiation) and need to generate a figure showing observed peaks, their intensities, and assigned fragment ions (e.g., b/y ions) for publication or presentation.

Use when when migrating spectral library data from file-based formats (JSON, CSV, binary) into a persistent store and need to support fast filtered queries on metadata and similarity computations against query spectra.

Use when when you have mass spectrometry data (MS1/MS2 scans from ThermoFisher .raw files, mzML, or MGF) from cells or organisms cultured with stable isotope-enriched substrates (e.g., 13C, 15N, 2H) at any enrichment level (natural 1.

Use when you have XCMS-processed LC/MS peak tables from paired unlabeled (12C) and labeled (13C) metabolic samples with replicate measurements, and you want to systematically detect which features show significant enrichment in the labeled condition relative to the unlabeled control.

Use when you have a high-resolution mass spectrum (FT scan) containing a peptide precursor at known charge state with known or suspected stable isotope labeling (e.

Use when you have cloned the GNPS_MASST codebase and need to instantiate a domain-specific MASST variant (microbeMASST, plantMASST, tissueMASST, microbiomeMASST, or foodMASST) to accept individual MS/MS spectra as input queries and perform searches against the corresponding curated reference.

Use when you have ranked GCF-MF (Gene Cluster Family–Molecular Family) links using two or more independent scoring functions and a set of experimentally validated links.

Use when when you have peak area tables (unlabeled C12 and labeled C13) from LC-MS metabolomics with sample metadata indicating case and control groups, and you need to distinguish true metabolic changes from instrumental noise or batch artifacts before attempting isotopic pairing.

Use when you have computed raw or standardised correlation scores (or other link-ranking metrics) for all possible GCF-MF pairs in a dataset and want to verify that validated links (those with known strain co-occurrence or experimental confirmation) are significantly enriched at higher score values.