TFE supports both vendor-native and open LC-MS file formats for targeted and non-targeted workflows. The software is designed primarily for datasets acquired using consistent acquisition methods across all samples.

If you encounter file compatibility issues, please contact us at metabolomicssoftware@broadinstitute.org or open an issue in this repository.

System Requirements

TFE has no formal system requirements, but it requires roughly 1 GB of memory to load the program.

Additional memory usage depends on the number of samples, number of annotations, extraction range, and scan rate. A rough estimate is:

NumSamples × NumAnnotations × (Extraction Range × scan rate) × 2 × 8 bytes

where 2 represents the two stored values per point: intensity and either RT for QQQ data or m/z for HRAM data.

At 8 Hz, with 10,000 samples, 100 annotations, and ±1 minute XICs, TFE would require approximately 15.4 GB of RAM:

10,000 samples × 100 annotations × (2 min × 60 s/min × 8 Hz) × 2 values × 8 bytes = 15.4 GB RAM

This requirement may decrease in the future as lossless compression is implemented, such as zero-run-length encoding.

Performance and snappiness may slow when more files are added. On a modern system, switching metabolites with ~300 samples takes ~.3 seconds. At 3,000 it may take several seconds to switch.

Supported File Formats

Thermo RAW Files

TFE supports Thermo-Finnigan .raw files.

The software has been tested with data generated from instruments including:

  • Q-Exactive
  • Q-Exactive Plus
  • Exploris 240

Compatibility with additional Thermo instrument models may vary depending on acquisition settings and vendor format differences.

mzML Files

TFE also supports .mzML files.

Supported structures include:

  • ScanList (commonly used for full-scan MS data)
  • SpectrumList (commonly used for QQQ datasets)

.mzML files may provide broader compatibility across instruments and acquisition platforms.

Project and Data Assumptions

TFE can process a wide range of LC-MS datasets, but several assumptions are made during extraction, smoothing, and integration workflows.

1. Consistent LC-MS Method Across Samples

All samples within a project are expected to be acquired using the same LC-MS method.

If acquisition methods differ significantly, retention times may shift substantially between samples, which can cause automated peak extraction and integration workflows to fail. Manual integration and review can still be performed in these cases.

2. Relatively Uniform Scan Spacing

TFE assumes that scan spacing is relatively even across the chromatogram.

Highly irregular scan timing — commonly caused by aggressive data-dependent MS2 acquisition — can interfere with gaussian smoothing and affect peak integration.

3. No Polarity Switching

Positive/negative polairty switching acquisition methods are currently not supported. Support for polarity switching workflows may be added in a future release.