Claude Skills by HolobiomicsLab

Use when you have two or more complementary scoring functions (e.g., strain correlation and IOKR scores) that you wish to combine, and you need to determine which combination strategy and parameters maximize enrichment of known true links in a validation set.

Use when after implementing or modifying the scoring module that computes average InChIKey scores and neighbourhood scores for candidate matches, or when integrating new scoring logic into an existing MS2Query pipeline.

Use when you have executed batch searches across two or more domain-specific MASST tools and obtained separate output files (_microbe.json, _plant.json, _tissue.

Use when when you have completed a DIA-MS proteomics search (e.

Use when you have loaded DIA mass spectrometry search results containing feature identification data with associated Q-value scores, and you need to restrict the analytes available in selection drop-downs to those meeting a quality cutoff (typically 1% FDR equivalent) before visualization or manual.

Use when you have search result files from one or more DIA-MS analysis tools and need to load them into a unified environment for Q-value filtering, cross-tool comparison (upset plots), and interactive visualization of identifications, quantifications, and coefficient of variation metrics across.

Use when after peak detection on composite mass tracks when you need to evaluate whether a detected peak represents a pure, interference-free signal on its m/z channel.

Use when you have a trained neural network model and a labelled validation dataset (with high-quality and low-quality peak annotations), and you need to determine the optimal probability threshold that maximizes the difference between true positive rate and false positive rate for classifying MS1.

Use when you have labeled MS/MS spectra from replicate measurements and need to determine a frequency threshold for denoising that balances competing objectives: retaining true fragment signals while removing noise.

Use when you have extracted retention times from MS1 spectra for top signals in a single LC-MS/MS run and need to evaluate whether that gradient's separation performance is sufficient, or when you are building the objective function for an iterative gradient optimization loop where each candidate.

Use when you have extracted retention times from top MS1 features across an LC-MS run and need a single, comparable metric to evaluate how effectively a gradient spreads compounds across the chromatographic window.

Use when when evaluating whether an MS data processing platform (such as mzmine) supports the full range of separation/ionization techniques your laboratory uses, or when assessing whether gaps exist in the software architecture that would require external pre- or post-processing for specific.

Use when when you have raw LC-HRMS metabolomics data in .mzML or .

Use when after you have processed raw LC-MS/MS spectral data through the specXplore importing pipeline in a Jupyter notebook and produced an in-memory specXplore session data object containing t-SNE embeddings (based on ms2deepscore similarity scores) and associated spectral metadata.

Use when after computing activity scores for a collection of metabolite sets (pathways, GNPS Molecular Families, or MS2LDA Mass2Motifs) from intensity and annotation data.

Use when a scientific application (such as QCxMS2) requires multiple external programs with specific version constraints and you need to programmatically verify their presence and compatibility before executing calculations.

Use when you have developed an R-based workflow (e.g., data processing, peak detection, quality review) that is complex enough to warrant interactive parameter tuning and visual feedback, and you need to distribute it to collaborators or end-users who prefer a graphical interface over scripting.

Use when when refactoring a mass-spectrometry formula-prediction codebase that has deprecated a monolithic scoring function (FDRNet) and requires a modular, symmetric Siamese design to independently embed spectrum and molecular-formula features before combining them.

Use when after a TCN-based formula prediction model has generated initial formula candidates from MS/MS spectra, apply this skill to rescore and refine those candidates when you need to improve ranking accuracy.

Use when when you need to reduce overfitting in a Siamese neural network trained on mass spectrometry spectral pairs by adding weight regularization, or when users require flexible control over L1 and L2 penalty coefficients rather than hard-coded defaults.

Use when you have preprocessed MS/MS spectra binned into 10,000 equally-sized m/z bins (10–1000 m/z range) with square-root-transformed intensities, and you need to generate 200-dimensional spectral embeddings for structural similarity prediction, visualization via dimensionality reduction (e.

Use when you have a collection of preprocessed tandem mass spectra (binned into 10,000 equally-sized m/z bins, intensities square-root transformed, top 1,000 peaks retained), a trained MS2DeepScore Siamese model, and you need to predict structural similarity scores (Tanimoto or Dice) for all or a.

Use when you have a collection of cleaned MS/MS spectra (in formats like mzML, mgf, msp, mzxml, or json) and need to predict molecular structural similarities between spectrum pairs without pre-computing RDKit fingerprints.

Use when you have processed the same GC–MS dataset (m/z vs retention time) through two independent signal acquisition pipelines and need to compute and compare their detection accuracy. Use this skill when one method (e.

Use when when processing raw Bruker Solarix transient files (.d format) destined for FT-MS analysis, especially for ESI-negative or low-abundance natural organic matter samples where baseline noise and side-lobe artifacts around intense peaks degrade peak picking and formula assignment accuracy.

Use when after loading raw LC-MS data from multiple disease groups when you need to compute correlations between metabolite signals and disease classes, or before training a deep learning model on metabolomics profiles.

Use when after generating consensus spectra with fragment recurrence frequencies, when you have replicate MS/MS spectra for features and need to choose a single frequency cutoff for denoising. Triggers include: (1) uncertainty about which frequency threshold to apply across all features;

Use when analyzing 1D signal arrays (e.g., extracted ion chromatograms, arrival time distributions, or MS1 spectra intensity profiles) where multiple peaks may overlap or where peak shape information (amplitude, position, width) is required beyond simple local-maximum detection.

Use when when working with raw LC-HRMS profile-mode data containing noisy chromatographic signals, apply this skill before peak detection. Smoothing is particularly needed when the signal-to-noise ratio is low or when gradient-based peak detection would be compromised by high-frequency noise.

Use when after elution peaks have been detected on composite mass tracks using local maxima and prominence detection, before reporting features in the final feature table.

Use when after peak detection in nontargeted LC-MS workflows when you have a feature table with detected peaks and need to assign quality scores or filter low-confidence features.

Use when you have QCpool (pooled quality control) samples measured at regular intervals across one or more LC-MS/MS sequences and need to detect whether instrument performance degrades, drifts, or destabilizes during the analytical run.

Use when when you have a pre-computed hierarchical dendrogram from structural clustering (e.g., of LC-MS features based on m/z and retention time) and want to compare or validate the cluster assignments produced by a fixed constant-threshold method.

Use when when you have cleaned and filtered mass spectrometry spectral data (in mzML, mzXML, msp, MGF, or JSON format) and need to identify or rank spectra by similarity for library matching, metabolite annotation, or network analysis.

Use when you have a collection of cleaned spectra in supported formats (mzML, mzXML, msp, MGF, JSON) and need to compute all pairwise similarity scores to identify related or duplicate spectra, support spectral library searching, or enable network-based analysis.

Use when after training contrastive embeddings that unify MS/MS spectra and molecular structures into a shared embedding space.

Use when after a deep-learning model has predicted structural similarity scores between an unknown metabolite's MS/MS spectrum and all known metabolites in a reference database.

Use when you have LC-MS/MS query spectra in mgf format that you need to match against a custom database (e.g., prepared with CFM-id) to identify compounds. Apply this skill when you want to rank candidate compounds by spectral similarity and return scored match results for downstream interpretation.

Use when when you have computed Spec2Vec similarity scores (typically cosine similarity in [0, 1] range) between discovered Mass2Motifs and a spectral library, and need to decide which matches are sufficiently confident to include in per-motif annotation output.

Use when you have a set of metabolites or chemical formulas to analyze and want to evaluate how different MS/MS fragmentation strategies (e.g., TopN, exclusion lists, dynamic window selection) would perform without access to real instrument time.

Use when when you have simulated DDA (data-dependent acquisition) scans from a ViMMS Environment and need to (1) quantify how well the simulated acquisition matched real or reference data (via evaluation metrics), and (2) export the results as standards-compliant mzML files for comparison with.

Use when after a ViMMS Environment.run() simulation completes with save_eval flag enabled, you have collected EvaluationData containing chemical compounds, their generated scans, and fragmentation events in memory.

Use when you have detected monoisotopic features (m/z, drift_time, retention_time, intensity) from LC-IMS-MS/MS data and need to identify and cluster their C13 isotopologues for charge state z=+1.

Use when when you have a log2-normalized, zero-mean, unit-variance intensity matrix (rows=metabolites, columns=samples) and a curated metabolite set database (e.

Use when your LC-HRMS metabolomics analysis must run on a high-performance computing cluster (e.g., HiPerGator, SLURM-managed systems) that lacks Docker support or prefers Singularity for security and portability. You have .mzML or .

Use when you have a Nextflow workflow (e.g., Nextflow4MS-DIAL) that currently runs under Docker or bare metal, and you need to execute it on an HPC cluster that lacks Docker support or enforces Singularity as the container runtime.

Use when you have variable-length lists of MS/MS peaks (m/z and intensity pairs) that need to be encoded into a fixed-dimensional representation compatible with transformer architecture.

Use when when you have processed LC-MS/MS data with precursor m/z, ionization mode, collision energy (if available), and fragment peak lists (m/z and intensity pairs), and need to query CSI:FingerID for molecular fingerprint predictions as part of an automated metabolite identification workflow.

Use when you have computationally generated precursor m/z values, fragment m/z values, collision energies, and retention time predictions for a set of lipid targets (e.

Use when rapid QC-MS detects a QC failure (e.g., internal standard retention time drift, m/z deviation, or intensity anomaly) during an active LC-MS instrument run and users have configured Slack as a notification target.