AI & ML

holobiomicslab Skill

Use when after successfully training a spectrum prediction model (FFN encoder, GNN encoder, intensity predictor, or fragment generator) to completion or at intermediate milestones, and before using that model for inference on test sets, structural elucidation queries, or transfer learning.

holobiomicslab Skill

Use when when you have access to annotated MS/MS spectra from a specific ionization mode (e.g., negative ESI) or adduct class (e.

holobiomicslab Skill

Use when after fitting a Multi-Block PLS (MB-PLS) discriminant or regression model on multi-assay LC-MS intensity data (e.g., HPOS, LPOS, LNEG blocks), and you need to identify which features drive model performance and warrant further statistical validation or biological interpretation.

holobiomicslab Skill

Use when you have positive-mode tune mix reference data (e.g., example_tune_pos.h5) with known CCS values spanning a wide m/z range (e.g., 118.086–1522 m/z) and need to establish a CCS calibration model to convert experimental drift times or collision cross sections for downstream analysis.

holobiomicslab Skill

Use when you have raw spectra files from a liquid chromatography experiment (in-house or external database) and need to adapt a pretrained GNN-RT model to predict retention times for your molecules. Preprocessing is the mandatory first step before any model training or transfer learning can proceed.

holobiomicslab Skill

Use when you have a collection of molecular structures (SMILES or chemical graphs) with paired experimental tandem mass spectra and want to build or benchmark a predictive model.

holobiomicslab Skill

Use when after training or loading a NeatMS neural network model, before applying it to filter false positive MS1 peaks in a new dataset.

holobiomicslab Skill

Use when after biomolecular class labels have been assigned to features in a TWIM-MS dataset and you have raw ion mobility arrival time measurements. Use it when you need to convert arrival times to standardized CCS values where calibration accuracy depends critically on the biomolecular class (e.

holobiomicslab Skill

Use when when you have executed multiple clustering tools on the same tandem-MS dataset and need to quantitatively compare their performance using normalized, comparable metrics rather than raw cluster assignments alone.

holobiomicslab Skill

Use when you have preprocessed MS/MS spectral data (normalized peak intensities and m/z values) in memory or on disk, a trained CNN model checkpoint available, and you need to generate molecular embedding vectors for matching against a reference database of known metabolites.

holobiomicslab Skill

Use when you have two or more feature tables in HDF5 format with detected features characterized by m/z, drift time, retention time, and intensity, and you need to match corresponding features across samples to account for systematic shifts caused by instrument variation or tuning differences.

holobiomicslab Skill

Use when when training a Siamese architecture rescore model for MS/MS-based molecular formula prediction and you have an imbalanced training set with far fewer negative than positive spectrum pairs.

holobiomicslab Skill

Use when after training a Siamese neural network on MS/MS spectrum pairs, use this skill to quantify prediction performance on a disjoint test set (e.g., 3600+ spectra from 500 unseen compounds).

holobiomicslab Skill

Use when after integratePeaks has been executed with a chosen integration method (e.g., fixed_size with RIP saturation threshold of 0.1) on a clustered, baseline-corrected GC-IMS dataset.

holobiomicslab Skill

Use when you have a large, mixed-instrument GNPS spectral dataset and need to create an instrument-specific training set for FIDDLE or similar deep learning models. Use this skill when your configuration file specifies an instrument allowlist (e.

holobiomicslab Skill

Use when you have pre-trained Keras models from the NP-Classifier repository that must be deployed via TensorFlow Serving and need to expose standardized input/output layer names ('input_2048', 'input_4096', 'output') for integration with the classification API.

holobiomicslab Skill

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), ...

holobiomicslab Skill

Use when after generating tandem mass spectrum predictions from a neural model (ICEBERG, SCARF, or baseline), and before attempting retrieval ranking, metric computation, or validation against experimental spectra.

holobiomicslab Skill

Use when after running inference on a trained structure prediction model with one or more input modalities (1H NMR, 13C NMR, or combined), you have generated predicted molecular formulas and connectivity graphs that need to be compared against known ground truth structures.

holobiomicslab Skill

Use when you need to measure how much a specific model capability or architectural feature contributes to prediction performance, especially when that capability is non-obvious or orthogonal to baseline methods.

holobiomicslab Skill

Use when after executing a molecular networking workflow on GC-MS data that has been processed through auto-deconvolution, and a published reference network exists from a prior analysis of the same or analogous dataset.

holobiomicslab Skill

Use when you have loaded an MsmsSpectrum object from a proteomics or metabolomics dataset and need to focus the analysis window on a specific m/z range relevant to your experiment (e.g., 100–1400 m/z for typical tryptic peptides).

holobiomicslab Skill

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.

holobiomicslab Skill

Use when after training or evaluating a classification model (e.g., MS2DeepScore or other neural networks) when you have paired arrays of predicted class labels and ground-truth labels and need to assess per-class prediction accuracy, false positive/negative rates, or class imbalance effects.

holobiomicslab Skill

Use when after instantiating and invoking a sinusoidal formula embedding layer (such as SCARF embeddings in MIST-CF) on chemical formula inputs, validate that the output embeddings meet dimensionality and value constraints before using them for downstream transformer or ranking tasks.

holobiomicslab Skill

Use when when preparing MS/MS spectral training data where: (1) the initial TCN train/test split contains imbalanced positive and negative examples, (2) certain molecular formulas are over-represented in the positive class, (3) you are training a Siamese architecture rescore model that requires.

holobiomicslab Skill

Use when after a trained CNN model has generated molecular embeddings for query spectra, and you need to retrieve the most likely candidate molecules from a reference database.

holobiomicslab Skill

Use when when you have a cleaned MS/MS dataset with chemical structure annotations (SMILES, InChI, or InChIKey) and need to generate ground-truth structural similarity labels for training a deep learning model.

holobiomicslab Skill

Use when when you have a molecular target compound defined by SMILES, InChI, or chemical formula and need to feed it into a pretrained spectrum prediction model (ICEBERG or SCARF) to generate tandem mass spectra or conduct structural elucidation.

holobiomicslab Skill

Use when you have a USI string referencing a spectrum in an online public repository (PRIDE, MassIVE, etc.) and need to load its raw spectral data without downloading the entire dataset file.

holobiomicslab Skill

Use when you have a dataset of SMILES strings with corresponding experimental CCS measurements and want to build a predictive model that can rapidly generate CCS values for new molecules without running expensive ion-mobility spectrometry experiments.

holobiomicslab Skill

Use when you have (1) a benchmark dataset of known molecules with accurate m/z values, retention time boundaries, and isotopologue identifiers for all enviPat-predicted adducts;

holobiomicslab Skill

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.

holobiomicslab Skill

Use when before submitting peak data or other inputs to a machine learning classification API for the first time, after a model update, or if you encounter unexpected prediction errors. It is essential when the underlying model's input names or structure may change and require code updates.

holobiomicslab Skill

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.

holobiomicslab Skill

Use when you have a formula-assigned dataset from FT-ICR MS (or other compound-annotated mass spectrometry) and you want to characterize molecular transformations and their co-occurrence patterns—particularly in studies of DOM reactivity, fermentation, or oxidative treatment of organic mixtures.