Chroma-Agent-Alpha: An Open-Source Framework for Automated Chromatography Preprocessing, GNN-Based Peak Deconvolution, and FAIR Data Lineage
- 1Department of Chemistry, Faculty of Engineering
- 2Teerthanker Mahaveer University
Abstract
Gas and liquid chromatography coupled with mass spectrometry (GC-MS/LC-MS) are foundational to analytical chemistry, quality control, and metabolomics. However, manual baseline removal, dynamic peak integration, and retention time alignment across multiple sample runs remain persistent bottlenecks. Commercial chromatography data systems depend on locked binary formats, manual integration tweaks, and inflexible rules. We introduce Chroma-Agent-Alpha, an open-source software framework that automates raw chromatography processing. The system implements native binary parsers for Agilent .ch and Varian .xms formats, Whittaker-Eilers Asymmetric Least Squares (ALS) baseline correction, Sakoe-Chiba constrained Dynamic Time Warping (DTW), and a two-layer Graph Convolutional Network (GCN) paired with Exponentially Modified Gaussian (EMG) non-linear fitting to deconvolve co-eluting analyte signals. Output matrices are formatted into Zarr v3 coordinate arrays and indexed in a local SQLite LaminDB database for FAIR data lineage tracking. We benchmarked the GNN model against the 2023 PeakDetective baseline. On ambiguous co-eluting peaks, the curation F1-score rose from 0.2222 to 0.9787, while peak integration error dropped by 48.8% in under 12 seconds on standard CPU hardware.
Related
- Deepweed: automated individual segmentation and tracking for duckweed lineage analysis
- WDGTA: a wavelet-denoised graph attention framework for spatial transcriptomics deconvolution.
- A unified and composable framework for modeling packed-bed liquid chromatography
- Hydrophobic Interaction Chromatography Framework for Drug-to-Antibody Ratio Assessment of a High DAR Cysteine-Linked ADC
- GenAI-Net: A generative AI framework for automated biomolecular network design
BibTeX
@misc{kataria2026chromaagentalpha,
title = {{Chroma-Agent-Alpha: An Open-Source Framework for Automated Chromatography Preprocessing, GNN-Based Peak Deconvolution, and FAIR Data Lineage}},
author = {Devendra Kataria},
howpublished = {ChemRxiv},
year = {2026},
month = aug,
doi = {10.26434/chemrxiv.15008097/v1},
url = {https://chemrxiv.org/doi/full/10.26434/chemrxiv.15008097/v1},
note = {CC BY},
}
RIS
TY - UNPB
TI - Chroma-Agent-Alpha: An Open-Source Framework for Automated Chromatography Preprocessing, GNN-Based Peak Deconvolution, and FAIR Data Lineage
AU - Devendra Kataria
PY - 2026
DA - 2026/08/31
PB - ChemRxiv
DO - 10.26434/chemrxiv.15008097/v1
UR - https://chemrxiv.org/doi/full/10.26434/chemrxiv.15008097/v1
AB - Gas and liquid chromatography coupled with mass spectrometry (GC-MS/LC-MS) are foundational to analytical chemistry, quality control, and metabolomics. However, manual baseline removal, dynamic peak integration, and retention time alignment across multiple sample runs remain persistent bottlenecks. Commercial chromatography data systems depend on locked binary formats, manual integration tweaks, and inflexible rules. We introduce Chroma-Agent-Alpha, an open-source software framework that automates raw chromatography processing. The system implements native binary parsers for Agilent .ch and Varian .xms formats, Whittaker-Eilers Asymmetric Least Squares (ALS) baseline correction, Sakoe-Chiba constrained Dynamic Time Warping (DTW), and a two-layer Graph Convolutional Network (GCN) paired with Exponentially Modified Gaussian (EMG) non-linear fitting to deconvolve co-eluting analyte signals. Output matrices are formatted into Zarr v3 coordinate arrays and indexed in a local SQLite LaminDB database for FAIR data lineage tracking. We benchmarked the GNN model against the 2023 PeakDetective baseline. On ambiguous co-eluting peaks, the curation F1-score rose from 0.2222 to 0.9787, while peak integration error dropped by 48.8% in under 12 seconds on standard CPU hardware.
ER -