AI & ML

holobiomicslab Skill

Use when after training a GNN model on molecular structures with continuous targets (e.g., CCS values), when you need to understand which node-level (atom) or edge-level (bond) features contribute most to individual or aggregate predictions.

holobiomicslab Skill

Use when you have downloaded raw spectroscopic data files (NMR, HSQC, COSY, IR modalities) from the Zenodo repositories and need to convert them into the standardized multi-modal input format required by the MultiModalSpectralTransformer before inference or retraining.

holobiomicslab Skill

Use when after executing a molecular structure prediction model on spectroscopic input data and obtaining predicted molecular structures in a standardized format (e.g., SMILES, MOL, SDF).

holobiomicslab Skill

Use when after an NMR-based structure prediction model has generated predicted molecular structures (formula and connectivity) for a test set of molecules with up to 19 heavy atoms.

holobiomicslab Skill

Use when you have received spectrum predictions (fragment masses and intensities) from a neural model (ICEBERG, SCARF, or similar) and need to extract structural information, rank candidate molecules, or validate predictions against experimental spectra.

holobiomicslab Skill

Use when preparing chemical formulae (e.g., 'C6H12O6') as inputs to a transformer-based neural network for MS/MS spectrum scoring. Use it specifically when the transformer must rank multiple candidate formulae against an observed mass spectrum and element composition constraints are important;

holobiomicslab Skill

Use when you have a dataset of molecules with known or reference CCS values, and you need to construct a trainable model that learns the mapping from molecular structure (encoded as SMILES or feature vectors) to scalar CCS predictions.

holobiomicslab Skill

Use when you have preprocessed mass spectrometry spectra (tokenized m/z and intensity pairs or feature matrices) and a trained deep learning model checkpoint, and you need to classify unknown compounds or generate prediction confidence scores for structural novelty analysis.

holobiomicslab Skill

Use when after a CNN model has generated predicted molecular embeddings from mass spectrometry data, and you need to identify the most likely candidate molecules from a reference database.

holobiomicslab Skill

Use when you have a mass spectrum from an untargeted metabolomics experiment and a set of candidate molecules (e.g., downloaded from PubChem) that may explain that spectrum. You want to rank these candidates by likelihood of correctness to prioritize manual annotation or further validation.

holobiomicslab Skill

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.

holobiomicslab Skill

Use when before feeding a peak list into the NMRformer model or other transformer-based spectral assignment frameworks.

holobiomicslab Skill

Use when when you need to programmatically interface with a TensorFlow Serving model instance and must discover or validate the expected input names (e.

holobiomicslab Skill

Use when you have a trained multitask model checkpoint and preprocessed spectral inputs (1D NMR spectra, 1H-only, 13C-only, or combined 1H+13C), and you need to generate simultaneous predictions of molecular formula and connectivity structure to quantify modality contributions, compare single vs..

holobiomicslab Skill

Use when when you have a dataset of molecular structures encoded as SMILES strings that will be processed downstream (e.

holobiomicslab Skill

Use when when you have multi-modal spectroscopic data (IR, Raman, UV-Vis, mass spectra, or NMR) and need to recover the underlying molecular structure without relying on finite spectral libraries or autoregressive SMILES generation.

holobiomicslab Skill

Use when when you have a trained GNN model for molecular property prediction (e.g., collision cross section) and need to identify which graph structural features—atomic properties, bond types, or higher-order graph descriptors—are driving the model's predictions.

holobiomicslab Skill

Use when after a CNN-Transformer peak detection model has been run on LC-MS ROI images and has output predicted peak locations with confidence scores.

holobiomicslab Skill

Use when after hierarchical fragmentation library matching has produced candidate lipid annotations for a multi-species LC-MS/MS dataset, and you have applied retention time–based filtering rules (e.

holobiomicslab Skill

Use when after deploying a TensorFlow Serving container (especially within a Dockerized stack like NP-Classifier), before running classification or inference pipelines, to confirm that input layers are named 'input_2048' and 'input_4096' and output layer is named 'output'.

holobiomicslab Skill

Use when after LDA inference has produced a trained motifset (motifset.json or motifset_optimized.json) with Mass2Motif probability distributions over fragments and neutral losses.

holobiomicslab Skill

Use when you have a collection of MS/MS spectra in MGF format and need to prepare them for GPU-based clustering. Dataset size and available GPU memory are critical: use GTX 1080Ti for smaller proteome datasets; use GTX 3090 for datasets like PXD000561 that exceed GTX 1080Ti capacity.

holobiomicslab Skill

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.

holobiomicslab Skill

Use when you have three distinct mass spectrometry data sources (quantification table, metadata table, and spectral data from an MS library or reference dataset like omsw_pleurotus_ms2deepscore) and need to combine them into a single JSON output that preserves all three modalities for interactive.

holobiomicslab Skill

Use when when you have a collection of mass spectra (e.

holobiomicslab Skill

Use when your untargeted LC/HRMS dataset contains Data-Independent Acquisition data (MS^E, AIF, or SWATH-MS) or MS1-only composite spectra where multiple precursor ions fragment simultaneously, and you need to deconvolve overlapping fragmentation spectra to enable accurate chemical structure.

holobiomicslab Skill

Use when you have resolved USI (Unified Spectrum Identifier) spectrum data from a supported repository (GNPS, MassBank, MetaboLights, Metabolomics Workbench, ProteoXchange, or MS2LDA) and need to create a figure suitable for journal publication or supplementary materials that retains a link to the.

holobiomicslab Skill

Use when you have raw or semi-processed 1D NMR spectra (¹H and/or ¹³C) from routine laboratory instruments and need to feed them into a CNN–transformer architecture for end-to-end structure elucidation.

holobiomicslab Skill

Use when you have a metabolomics dataset with metabolite identifiers in mixed formats (e.g., common names, KEGG accessions, HMDB IDs) and need to assign each metabolite to its canonical pathway(s) before performing pathway-level classification, feature selection, or prognosis modeling.

holobiomicslab Skill

Use when after you have (1) extracted regions of interest (ROIs) around candidate peaks in LC-MS data and (2) run those ROIs through a trained CNN-Transformer peak detection model that outputs both binary peak classifications and bounding box coordinates.

holobiomicslab Skill

Use when you have loaded MSI data with an extracted peak list and need to annotate matrix-related signals, particularly when the dataset may contain isobaric ions or peaks with overlapping spatial distributions that could be misclassified during downstream annotation filtering.

holobiomicslab Skill

Use when after importing imzML or vendor-specific MSI data into napari and visualizing the raw spectral dataset.

holobiomicslab Skill

Use when you have two independent LC-MS untargeted metabolomic feature datasets (each with retention time and m/z values) and need to identify which features in one dataset correspond to features in the other.

holobiomicslab Skill

Use when you have nuclear magnetic resonance (NMR) peak data (proton 1H and carbon-13 13C measurements) that you need to classify using a deployed deep learning model, and you have access to a TensorFlow Serving instance running the SMART 3 classification model.

holobiomicslab Skill

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.

holobiomicslab Skill

Use when you have a trained sequence-to-sequence model (such as GCMSFormer) that predicts mass spectra from overlapped peaks, and you need to evaluate model performance on a held-out test set.