Claude Skills by HolobiomicsLab

Use when after completing metabolite annotation of LC-MS AIF features using the annotateRC function, when you need to persist ranked candidate matches, inspect multiple candidate annotations per feature, visualize matched ions in ranked spectra, or export pseudo-MS/MS spectra for external analysis.

Use when when you have raw PubChem compound records or other public chemical structure databases and need to supply candidate metabolite structures to a mass spectrometry annotation workflow like MAGMa.

Use when you have experimental mass spectra from untargeted metabolomics and need to assign compound identities with confidence estimates.

Use when after obtaining in silico annotations from SIRIUS (Zodiac/Cosmic scores) or ISDB (cosine/shared peaks metrics), before using the annotation table for Feature Component calculation, chemical class assignment, or metabolite discovery prioritization.

Use when you have a normalized LC-MS/MS metabolite abundance matrix (e.g., from MS-DIAL preprocessing) with multiple samples across experimental groups (e.

Use when when you have downloaded a GNPS molecular networking job archive (from GNPS1 or GNPS2 workflows: METABOLOMICS-SNETS, METABOLOMICS-SNETS-V2, FEATURE-BASED-MOLECULAR-NETWORKING, classical_networking_workflow, or feature_based_molecular_networking_workflow) and need to extract and standardize.

Use when you have region-of-interest (ROI) LC-MS data and a pre-trained object detection model has already predicted peak locations and confidence scores.

Use when when you have Parquet-encoded tabular data (e.

Use when after calling buildExperiment on metabolomics data to construct a SummarizedExperiment object, before performing batch correction, outlier detection, or other QC calculations that depend on ratio assay values.

Use when when you have pretrained encoder-produced embeddings from MS/MS spectra and need to decode them into canonical SMILES strings representing molecular structures.

Use when you have imaging mass spectrometry (IMS) datasets with peak intensity features organized as spatial graphs (nodes = pixels/voxels, edges = spatial adjacency), and you need to discover latent peak patterns for automatic peak picking or marker ion identification without manual feature.

Use when after instantiating a transformer encoder module for mass spectrometry data processing (e.g., in IDSL_MINT), before training on large MS/MS datasets or running inference on test spectra.

Use when when you have raw or processed direct-infusion MS (DI-MS) or ASAP-MS spectra as mz/intensity pairs and need to rapidly identify salient peaks for species authentication, sample scoring, or comparative profiling without manual inspection.

Use when you have high-resolution tandem mass spectrometry (MS2) data in .ms2 format and need to systematically identify and annotate lipid A molecular structures.

Use when you have ion mobility-mass spectrometry data (raw drift times, m/z values, and feature intensities) from DTIMS-MS or SLIM-based IMS-MS platforms and need to compute collision cross section values using a calibration standard (e.g., Agilent tune-mix).

Use when you have a two-dimensional GC–MS or LC–MS dataset (m/z vs retention time) and need to identify discriminative analyte features without relying on conventional peak picking algorithms. This is especially valuable when analyzing complex, low-abundance samples (e.

Use when when you have two or more MsBackend objects (e.g., multiple MsBackendMemory, MsBackendDataFrame, or MsBackendMzR instances) containing distinct sets of MS spectra and you need to consolidate them into a single backend for downstream analysis, parallel processing, or export.

Use when when you have trained predictive models (e.g., neural networks) on paired microbiome-metabolome data and need to identify which metabolites are genuinely well-predicted above chance.

Use when after generating an initial LC-MS feature table (from Asari or similar peak detection) and before normalization or statistical analysis, especially when blank samples (e.g., solvent-only or buffer-only runs) were acquired alongside study samples.

Use when you have a feature table from LC-MS data alongside blank (solvent-only) sample runs, and you want to remove features whose intensity in study samples is not substantially higher than their intensity in blanks.

Use when processing MS-DIAL peak lists from untargeted LC-MS/MS experiments (DDA or DIA mode) where you suspect instrumental background contamination or ion source carry-over is generating false positive features.

Use when after noise filtering and polarity selection of MS/MS spectra (from .mgf, .mzML, or .msp files), when you need to prepare spectra for unsupervised discovery of recurring fragmentation patterns without prior compound identification.

Use when you have raw MS/MS spectral data in standard mass spectrometry formats and need to prepare it for unsupervised substructure discovery via topic modeling.

Use when you suspect XCMS grouping contains misaligned features due to suboptimal parameter settings or insufficient samples.

Use when before peak detection on a composite or individual mass track when you need to filter out low-intensity noise and baseline drift without removing true signal.

Use when your research proposes a new spectral embedding, matching algorithm, or retrieval method and you need to quantify its improvement over known baselines. Specifically, when you have a test dataset (e.

Use when when you have trained a candidate model (e.g., an ensemble, a new architecture) and need to demonstrate its advantage over published or reference implementations on the same test data.

Use when when you have implemented a novel annotation algorithm or network and must verify it outperforms a simpler reference approach on held-out or independent spectral data. Specifically: (1) you have ground-truth annotations for a common test set;

Use when you have developed or adapted an analytical method (e.g., NPFimg for GC–MS marker identification) and need to demonstrate its reliability or improved performance over a widely-used reference method (e.g., XCMS). Apply this skill when you have access to both the same raw input data (e.

Use when after auditing and optionally rescaling a mass track (composite mass chromatogram) when you need to subtract background signal and set dynamic prominence thresholds for peak detection.

Use when your LC-MS feature table exhibits intensity variations across samples that correlate with technical batches (e.

Use when preparing labeled LC-MS peak data for neural network training and you need to decide whether class imbalance in your dataset should be preserved or corrected in batch construction. Use it particularly when your annotated peak dataset has unequal class distributions (e.

Use when after applying batch correction (e.g., ComBat, SVA) to a merged feature table from non-targeted LC-MS/MS metabolomics data and before proceeding to univariate or multivariate statistical analysis.

Use when after applying pycombat-based batch correction to multi-batch interpolated feature tables in LC-MS metabolomics workflows, when you need to verify that batch effects have been attenuated without loss of data dimensionality or sample information.

Use when when you have processed metabolomics LC-MS/MS data organized by batch and sample type (including pooled QC replicates), and you need to quantify whether batch-to-batch and matrix effects are acceptable for downstream reporting.

Use when you have a feature table generated from LC-MS/MS non-targeted metabolomics data that spans multiple sample preparation batches, instrumental runs, or experimental conditions.

Use when after merging feature tables from multiple LC-MS/MS analytical runs or sample cohorts processed in separate batches, and before applying batch correction.

Use when you have raw mzML files and a feature table (CSV) from LCMS data processed by tools like mzMine, and you need to create train/test/validation batches with specific matrix dimensions (120 × 2) and verified margin/peak signal separation before training or evaluating a neural network.

Use when apply batch normalization after dense hidden layers (but not the final embedding layer) in a deep neural network trained on MS/MS spectral data, particularly when the network processes high-dimensional binned spectra (9948-dimensional vectors) and you need to stabilize gradient flow across.

Use when when you have raw mzML files and a feature table (CSV from mzMine or XCMS) with labeled peaks of unequal class sizes (e.g., fewer false positives than true positives) and plan to train a CNN classifier on the LCMS data.

Use when you have acquired MS/MS spectra in .msp format (e.g., from MassBank or experimental acquisition) and need to transform them into a structured library format compatible with automated annotation tools.

Use when when processing a SummarizedExperiment containing metabolomics peak areas or intensities organized by batch (e.g., LC-MS/MS runs on different days or instruments), and you have pooled quality control samples injected repeatedly within or across batches to serve as batch anchors.

Use when you have LC/MS feature data (m/z, retention time, intensity) and need to assign metabolite annotations with confidence scores rather than binary peak-to-compound matches.

Use when you have experimental retention times (RTs) for a small set of calibration molecules (≥10) measured on both a source chromatographic method and a target method, and you need to predict RTs for candidate metabolites on the target method to rank annotation candidates.

Use when you have completed one or more LC-MS gradient runs, extracted separation efficiency metrics from the resulting MS1 and MS2 spectra, and need to incorporate those real experimental observations into your Gaussian process model to improve the next gradient proposal.

Use when after fitting a Gaussian Process regression model to prior LC-MS gradient evaluations (where gradients are encoded as input and separation efficiency is output), use this skill to decide which candidate gradient to test next.

Use when using Casanovo for de novo peptide sequencing on high-stakes datasets (immunopeptidomics, paleoproteomics, or monoclonal antibody discovery) where missing the correct sequence in the top-1 prediction is costly.

Use when you have post-processed clustering results from multiple tools (msCluster, Falcon, MaRaCluster) on the same tandem MS dataset and need to generate a comparative performance report with quality metrics and runtime statistics to determine which tool suits your metabolomics workflow.

Use when you have trained a new machine learning model for chemical formula or adduct assignment from MS/MS spectra and need to assess whether it offers genuine performance gains over established baselines. Use it specifically when you have access to ground-truth annotations (e.

Use when you have assembled genome sequences (contigs or scaffolds in FASTA format) and want to identify putative BGCs and their precursor peptides before constructing a RiPP structure database for spectral matching.