DIA-NN EasyFilter workflow for the fast and user-friendly critical assessment and visualization of DIA-NN proteomics analysis outcome

To address the difficulty of interrogating DIA-NN's compressed PARQUET output without programming skills, the authors developed DIA-NN EasyFilter (DEF), a fast and user-friendly KNIME-based workflow that enables comprehensive protein filtering, quality assessment, and interactive visualization for non-coders.

The Gist

Imagine you are a detective trying to solve a massive crime scene. The crime scene is a drop of blood or a piece of tissue, and the "suspects" are thousands of different proteins. To find out who is there, you use a high-tech machine (a mass spectrometer) that takes a picture of every single molecule.

The problem? The machine produces a photo that is so huge, so compressed, and so full of tiny details that it looks like a giant, unreadable spreadsheet of alien code. This is what the popular software DIA-NN does. It's incredibly smart and finds the suspects, but it hands you the evidence in a locked, compressed file (called a PARQUET file) that requires you to be a computer programmer to unlock and make sense of it.

Enter the "DIA-NN EasyFilter" (DEF).

Think of DEF as a friendly, user-friendly "Evidence Sorting Robot" built on a platform called KNIME. You don't need to know how to code to use it; you just drag and drop blocks like a child playing with LEGO.

Here is how the paper explains this tool using simple analogies:

1. The Problem: The "Locked Suitcase"

The original DIA-NN software is like a brilliant detective who puts all the evidence into a locked, heavy suitcase (the PARQUET file). If you aren't a locksmith (a programmer), you can't open it to see what's inside. You might know that there are suspects, but you can't easily check if they are real or just fake alibis (contaminants).

2. The Solution: The "Magic Filter"

The authors built DEF, which is like a magic sieve. You pour the heavy suitcase of data into the top, and DEF sorts it out for you instantly.

• It checks the ID cards: It looks at the "suspects" (proteins) and asks, "Do you have at least two unique ID cards (peptides) to prove you belong here?" If not, it kicks them out.
• It catches the imposters: Every crime scene has dust bunnies and random trash (contaminants from the lab, like human skin cells or plastic). DEF has a pre-loaded "Wanted List" of these imposters. It scans the evidence and instantly throws the trash away so you only see the real suspects.
• It checks the fingerprints: If you took extra photos (called XICs), DEF looks at the fingerprints to make sure the suspect was actually there and not just a blurry shadow.

3. The Dashboard: The "Control Panel"

Once the data is sorted, DEF doesn't just give you a boring list. It gives you a colorful dashboard with charts and graphs.

• The Pie Chart: Imagine a pizza. DEF shows you a pizza where the biggest slices are the most common proteins. It even shows you a tiny slice for the "trash" so you can see how much of your sample was actually junk.
• The Parallel Lines: Imagine a bunch of strings connecting different days of the experiment. If a string is broken (missing data), you can see it instantly. This helps you spot if your experiment was shaky or solid.

4. The Test Drive: "Case Studies"

The authors didn't just build the robot; they tested it on four different "crime scenes" (datasets) to prove it works:

• Case 1 & 2: They took data from other scientists' studies and ran it through DEF. The results matched perfectly, proving DEF is just as smart as the experts, but much faster and easier to use.
• Case 3: They tested it on data from different types of machines (Sciex instruments). DEF handled them all without breaking a sweat.
• Case 4 (The Real Deal): The authors used DEF on their own new experiment involving fat cells (adipocytes). They wanted to see how fat cells change when you feed them a high-fat diet (palmitate).
  • The Result: DEF helped them quickly find that the fat cells changed their "metabolism" (how they burn energy) and their "skeleton" (structure) when exposed to fat. It found specific proteins that acted like switches, turning certain processes on or off.

Why Does This Matter?

Before this tool, if you wanted to analyze this kind of data, you had to be a coding wizard or hire one. That's expensive and slow.

DEF is like giving a power tool to a regular homeowner.

• Speed: It can process a huge dataset (35 samples) in under 14 minutes.
• Simplicity: You click buttons instead of writing code.
• Trust: It filters out the noise so you can trust your results.

In a nutshell: The paper introduces a tool that takes the complex, locked-up results of a super-smart protein analyzer and turns them into a clear, colorful, and easy-to-read story, allowing scientists who aren't programmers to do high-level detective work on their own.

Technical Summary

1. Problem Statement

• Context: Data-Independent Acquisition (DIA) mass spectrometry, particularly using the DIA-NN software, has become a standard for quantitative proteomics due to its high accuracy and reproducibility.
• The Bottleneck: The primary output of DIA-NN (versions ≥1.8) is a PARQUET file. While efficient for storage, this compressed format is difficult to interrogate without programming expertise (e.g., Python/R with PyArrow).
• Limitations of Existing Solutions:
  • Script-based (R/Python): Requires coding skills, increasing the barrier for non-bioinformaticians and risking configuration errors.
  • Skyline: Powerful for quality assessment but requires significant manual effort and parametrization.
  • Galaxy: Good for reproducibility but less flexible for highly customized, modular desktop workflows compared to KNIME.
• Gap: There was a lack of a dedicated, user-friendly, low-code/no-code workflow specifically designed to filter, assess quality, and visualize DIA-NN reports (including PARQUET and XIC data) for end-users without programming skills.

2. Methodology: DIA-NN EasyFilter (DEF)

The authors developed DEF, a workflow built on the KNIME Analytics Platform (version 5.4.3), designed to run locally on desktop environments.

• Input Handling:
  • Accepts DIA-NN main outputs (PARQUET or TSV), pg.matrix files, and optional Extracted Ion Chromatogram (XIC) reports.
  • No external libraries are required beyond those native to KNIME.
• Core Filtering Modules:
  1. XIC-Based Filtration: If XIC data is available, the workflow extracts fragment ions (b- and y-ions). It retains only peptides supported by at least four consecutive b- or y-ions, significantly improving peptide identification confidence.
  2. Contaminant Filtering: Integrates curated contaminant lists from MaxQuant, cRAP, and the Hao-Group. Users can also upload custom FASTA files. Contaminants and iRT peptides are flagged and excluded.
  3. Protein Group (PG) Inference Strategies:
     • Two-Unique-Peptide Rule: Retains only PGs supported by >1 unique peptide sequence.
     • Proteotypic Rule: Retains PGs identified by proteotypic (protein-specific) peptides, allowing for single unique sequence identification if the peptide is unique to that protein.
  4. Quality Thresholds:
     • Applies run-specific PG.Q.Value (≤0.05) and library-level Lib.PG.Q.Value/Lib.Q.Value (≤0.01).
     • Handles missing quantification values (e.g., PG.MaxLFQ) by replacing non-quantified entries with a placeholder ("50") rather than excluding them entirely, allowing for visual distinction.
• Visualization & Output:
  • Generates interactive bar plots, stacked bar plots for quantification performance, and parallel coordinate plots (binary and scaled) to visualize identification consistency and abundance trends.
  • Includes a pie chart for intensity distribution (including contaminants).
  • Exports data in formats compatible with MetaboAnalyst for downstream statistical analysis.

3. Key Contributions

• Accessibility: Democratizes advanced DIA-NN post-processing for researchers with minimal programming background via a visual, node-based interface.
• Speed & Efficiency: Capable of processing large-scale reports (tens of millions of rows) in minutes. In Case 3, a 35-sample batch was processed in under 14 minutes.
• Integrated Quality Control: Uniquely combines chromatographic peak validation (XIC), contaminant removal, and flexible protein inference rules into a single automated pipeline.
• Reproducibility: Provides a standardized, traceable workflow that eliminates manual data handling errors common in script-based or manual inspection methods.

4. Results

The workflow was validated through four case studies using public datasets and in-house data:

• Case 1 (HEK293 Cells): Re-analysis of a public dataset showed DEF identified ~4,200–4,900 Protein Groups (PGs), comparable to the original publication. Quantification reproducibility (Median CV) was consistent (~6.6–6.9%), with slight improvements attributed to the use of DIA-NN v2.2 (QuantUMS algorithm).
• Case 2 (Mouse/Yeast Mix): Tested on Thermo QE HF and Bruker timsTOF Pro instruments. DEF results showed high similarity to the original study regarding the number of identified mouse proteins. DEF demonstrated slightly better reproducibility (Median CV 4.35–10.28% vs. 4.43–11.84% in the original).
• Case 3 (Tri-species Mixture - HYE124): Compared against OpenSWATH, SWATH 2.0, Skyline, Spectronaut, and DIA-Umpire.
  • DEF identified more proteins than all other tools (e.g., 4,873–6,707 proteins vs. 1,566–4,692 for others).
  • Demonstrated high technical reproducibility (Median CV 6–9%), comparable to top-tier tools like Spectronaut.
• Case 4 (Adipocyte Differentiation - SGBS): Applied to in-house data (Pre-adipocytes vs. Differentiated Adipocytes ± Palmitate).
  • Successfully identified differentially abundant proteins (DAPs) using both inference rules.
  • Enabled downstream statistical analysis in MetaboAnalyst, revealing biological insights:
    • Differentiation: Driven by cytoskeleton remodeling and oxidoreductase activity.
    • Palmitate Treatment: Triggered extracellular matrix remodeling (proteoglycans in cancer pathway) and metabolic shifts toward lipogenesis.
  • Identified specific proteins (e.g., MT-CO2, TF, RPL10) with altered abundance.

5. Significance

• Bridging the Gap: DEF fills a critical gap in the proteomics community by providing a "no-code" solution for the complex post-processing of DIA-NN outputs, specifically addressing the difficulty of handling PARQUET files.
• One Health & Clinical Applicability: The workflow is optimized for the medium-to-large datasets typical in One Health and clinical studies, offering a balance between the scalability of web tools (Galaxy) and the flexibility of desktop tools.
• Enhanced Data Integrity: By integrating XIC-based validation and rigorous contaminant filtering, DEF improves the reliability of protein identification and quantification, reducing false positives.
• Future-Proofing: While currently limited to DIA-NN outputs, the modular KNIME architecture allows for future expansion to other input formats and quantification algorithms.

Conclusion: The DIA-NN EasyFilter (DEF) workflow is a robust, fast, and user-friendly tool that enables researchers to critically assess, filter, and visualize DIA-NN proteomics data without coding expertise, ensuring high-quality, reproducible results across diverse experimental platforms.