MassCube: a Python framework for end-to-end metabolomics data processing from raw files to phenotype classifiers

MS-based nontargeted metabolomics now involves thousands of data files across multiple assays and laboratories. Meanwhile, new mass spectrometers like Orbitrap Astral MS have increased raw data sizes eightfold, demanding more efficient processing. Conventional software, including MS-DIAL, MZmine3, and XCMS, struggles with false-positive peak detection, inefficient in-source fragment handling, and slow computation. MassCube, an open-source Python framework, addresses these challenges with advanced peak detection, adduct/in-source fragment recognition, and high-performance parallel computing. Leveraging array-based programming, MassCube enables scalable, high-throughput analysis for large metabolomics datasets. Systematic benchmarking shows MassCube outperforms leading tools in speed, accuracy, and robustness. MassCube’s peak detection was benchmarked against MS-DIAL, MZmine3, and XCMS using both synthetic and experimental LC-MS datasets. In synthetic benchmarking, MassCube achieved 96.5% accuracy, significantly outperforming MS-DIAL (85.4%), MZmine3 (88.4%), and XCMS (87.4%). Experimental benchmarking involved eight LC-MS datasets including human plasma, serum, urine, fecal material, mouse plasma, fruit fly, and plant extracts. A total of 722 manually labeled ion traces were evaluated, with MassCube demonstrating superior accuracy for double-peak (93.5%) and single-peak (97.3%) detection, outperforming MS-DIAL (63.0% and 89.8%), MZmine3 (30.4% and 81.8%), and XCMS (28.3% and 70.3%). Additionally, MassCube exhibited unmatched processing speed and scalability. On a 105 GB Orbitrap Astral dataset, it processed 636 files in 64 minutes on a MacBook M3 Pro (36 GB memory, 12 cores), whereas other software required 8–24 times longer. The framework effectively handled large datasets while maintaining a low memory footprint, enabling high-throughput metabolomics analysis on standard computing hardware. Its modular, object-oriented design also facilitates rapid integration of new algorithms, making it adaptable for evolving metabolomics workflows.