HXMS: a standardized file format for HX-MS data

This paper introduces HXMS, a standardized, lightweight file format that preserves full isotopic mass spectra and comprehensive experimental metadata for Hydrogen/Deuterium Exchange-Mass Spectrometry (HX-MS) data, along with PFLink, a Python tool to convert existing software outputs into this format to enable more quantitative analysis, better data sharing, and future machine learning applications.

The Gist

The Problem: The "Lost in Translation" Crisis in Protein Science

Imagine you are trying to understand how a complex machine (like a protein) moves and changes shape. Scientists use a technique called HX-MS (Hydrogen/Deuterium Exchange Mass Spectrometry). Think of this technique as a high-speed camera that takes thousands of photos of a protein as it dances in a liquid.

For a long time, scientists have been taking these photos, but they've been doing it in a very messy way:

1. Different Languages: Every software company (Thermo Fisher, Waters, Trajan, etc.) saves these photos in their own secret language. It's like one photographer saves photos as JPEGs, another as TIFFs, and a third as a weird code only their camera understands. You can't easily share them or compare them.
2. Summarizing the Story: Most scientists only save the "average" of the photo. Imagine watching a chaotic dance party and only writing down, "The average person was dancing at 50 BPM." You lose all the detail: who was dancing fast, who was slow, and if there were two different groups dancing to different songs at the same time. This "average" approach throws away a huge amount of useful information.

The Solution: Introducing "HXMS" (The Universal Translator)

The authors of this paper have created a new, standardized file format called HXMS.

The Analogy: Think of HXMS as the PDF of the protein world.

• Before: Everyone had their own weird document format.
• Now: HXMS is a universal, lightweight, and human-readable format that anyone can open, read, and understand, regardless of which camera (software) took the original data.

What makes HXMS special?

• It keeps the full picture: Instead of just the "average" dance move, it saves the entire mass spectrum (the full isotopic envelope). It's like saving the raw video footage instead of just a summary sentence. This allows scientists to see if a protein is behaving in two different ways at once (multimodal distributions), which was previously impossible to track easily.
• It includes the "Cheat Sheet": It automatically includes all the experimental details (temperature, pH, time) so you never have to guess how the data was collected.
• It tracks the "Mods": Proteins often have little tags attached to them (Post-Translational Modifications). HXMS has a special dictionary section to list exactly what these tags are and where they are, so no detail is lost.

The Tool: "PFLink" (The Universal Adapter)

Creating a new file format is useless if nobody can convert their old files into it. That's where PFLink comes in.

The Analogy: Think of PFLink as a universal power strip adapter.

• You have a device from the UK, one from the US, and one from Japan (different HX-MS software).
• PFLink takes the data from any of these "outlets" and instantly converts it into the standard HXMS "plug" so it fits into any new system.
• It works with the four most popular software programs used in labs today. If you have data from one of them, you can plug it into PFLink, and it spits out a perfect HXMS file.

Why Does This Matter?

1. Better Science: By keeping the full, detailed data (not just the average), scientists can do much more precise math. They can calculate the energy of protein movements with higher accuracy.
2. Sharing is Caring: Because the format is standardized, scientists can easily share their data with colleagues around the world without worrying about compatibility. It's like sending an email attachment that everyone can open.
3. Future-Proofing: This sets the stage for Artificial Intelligence (Machine Learning). AI needs huge amounts of clean, standardized data to learn. HXMS provides that clean data, allowing computers to eventually help discover new drugs or understand diseases better.
4. Transparency: The format includes a "MATCH" section that acts like a receipt. It shows exactly how the data was processed, so if something looks weird, scientists can trace it back to the original raw numbers without needing the expensive, proprietary software from the vendor.

The Bottom Line

The authors are saying: "We built a universal filing cabinet (HXMS) and a magic converter (PFLink) so that all the messy, scattered protein data from labs around the world can finally be organized, shared, and understood together. This will help us solve biological mysteries faster and more accurately."

Technical Summary

1. Problem Statement

Hydrogen/deuterium exchange-mass spectrometry (HX-MS) is a critical technique for analyzing protein conformational ensembles, yet the field suffers from significant data interoperability and information loss issues:

• Lack of Standardization: The proliferation of diverse instrumentation and proprietary software (e.g., BioPharma Finder, HDExaminer, DynamX, HDX Workbench) has resulted in inconsistent, non-standardized data formats. This makes data sharing, archiving, and cross-platform analysis difficult.
• Information Loss: Most current workflows and file formats rely on "mean deuteration" (centroid) representations of isotopic mass envelopes. This practice discards the full isotopic distribution data, leading to information loss, degeneracy, and an inability to perform high-resolution quantitative analysis or detect multimodal distributions.
• Data Volume vs. Usability: While raw mass spectrometry files contain full spectral data, they are often instrument-specific, massive in size, and incompatible with downstream analysis tools that require standardized inputs.

2. Methodology

The authors developed a two-pronged solution: a new standardized file format (HXMS) and a conversion software package (PFLink).

A. The HXMS File Format

HXMS is designed to be lightweight, human-readable, and scalable. It is structured into three primary sections:

1. Metadata Section: Captures essential experimental conditions (protein sequence, state, temperature, pH, D2O saturation) and allows for custom remarks.
2. Experimental Data Section: Represents the core HX-MS data. Key features include:
   • Full Isotopic Envelopes: Unlike centroid-only formats, this section stores the full mass envelope (normalized intensity values) for every peptide at every timepoint.
   • Multimodal Support: Uses a "MOD" column (A–Z) to label distinct populations within a single peptide, enabling the analysis of conformational heterogeneity.
   • Replicates and PTMs: Explicitly tracks experimental replicates ("REP") and Post-Translational Modifications ("PTM_ID").
   • Back-Exchange Controls: Includes fully deuterated samples (time = "inf") to facilitate consistent quantification and correction.
3. PTM Dictionary Section: A lookup table linking PTM IDs to specific modification details (e.g., phosphorylation sites) using absolute positions.
4. MATCH Section (Optional): A traceability layer linking processed timepoints to the original raw spectral evidence. It stores charge states, monoisotopic masses, retention times, and raw m/z-intensity pairs (separating envelope peaks and fine isotopologue structures) to ensure vendor-specific processing biases are transparent and debuggable.

B. The PFLink Software

PFLink is a Python package designed to bridge the gap between existing workflows and the HXMS format.

• Input Compatibility: It accepts exported data from major commercial and academic software (BioPharma Finder, HDExaminer, DynamX, HDX Workbench) and a custom CSV template.
• Conversion Logic:
  • It handles mean deuteration data from all supported sources.
  • For software supporting full spectral export (HDX Workbench, HDExaminer), PFLink extracts the full isotopic envelopes to populate the HXMS format.
  • It automatically calculates deuterium uptake by normalizing against zero-timepoints.
  • It generates the MATCH section when full spectral data is available, preserving traceability.

3. Key Contributions

• Standardization: Introduction of HXMS as a unified, community-accepted format that decouples data from specific vendor software.
• Data Preservation: The format retains the full isotopic mass envelopes, enabling the recovery of information lost in centroid-based representations.
• Advanced Analysis Support: The format natively supports multimodal distributions (conformational sub-populations), PTMs, and experimental replicates, which are often difficult to handle in legacy formats.
• Traceability: The inclusion of the MATCH section allows researchers to audit peak assignments and processing steps without needing the original vendor software, addressing the "black box" nature of proprietary data processing.
• Accessibility: PFLink is open-source and available via HuggingFace, lowering the barrier to entry for converting legacy data.

4. Results and Validation

• Implementation: The authors successfully converted data from four different software platforms into the HXMS format.
• Case Studies: Two datasets were processed to demonstrate the format's versatility:
  1. E. coli DHFR: Data in apo, methotrexate-bound, and trimethoprim-bound states. The HXMS files included uncentroided fine structures (MATCH section) derived from raw spectra.
  2. HSV-1 gB: Pre- and post-fusion states. These files successfully captured bimodal spectra, demonstrating the format's ability to handle complex conformational ensembles.
• Compatibility: The generated HXMS files are compatible with advanced quantitative analysis tools like PFNet and FEATHER, which require high-resolution data inputs.

5. Significance and Future Impact

• Quantitative Advancement: By preserving full isotopic envelopes, HXMS enables more rigorous quantitative treatments of HX-MS data, including the determination of high-resolution ensemble energies.
• Data Sharing and Reproducibility: The format facilitates easier data sharing among practitioners and improves reproducibility by making processing steps transparent via the MATCH section.
• Machine Learning Readiness: The standardized, structured nature of HXMS makes it ideal for training machine learning models, which require large, consistent datasets from diverse sources.
• Community Evolution: The authors position HXMS not as a replacement for raw data repositories (like ProteomeXchange) but as a complementary, lightweight standard for publication and analysis. They advocate for software vendors to adopt HXMS export capabilities to future-proof the field.

In summary, this work addresses a critical bottleneck in structural biology by providing a standardized, information-rich data format and the necessary tools to transition the HX-MS community from proprietary, lossy formats to an open, high-resolution ecosystem.