Summary
- Auto-inspector is a new Byosphere product that automates the Data Inspection step of a Byos project, using customizable rules and decision trees to populate comments, labels, and validation fields.
- Applied to a PTM analysis test project, Auto-inspector's grouping and filtering logic alone reduced the number of items in the peptide table by 80%
- The underlying decision-tree logic is fully customizable and is not limited to PTM analysis – equivalent Auto-inspector logic has already been built for SVA and disulfide-linked peptide analysis.
Introduction
Each analysis performed in Byosphere consists of three steps: Project Creation, Data Inspection, and Reporting. Project Creation and Reporting can both be automated. For example, all steps needed to populate the workflow file can be automated, and the resulting workflow is automatically submitted for project creation. Report generation and the transfer of data to third parties, such as an ELN, can likewise be automated without limitation, as long as the third-party software exposes an API to send and receive data.
Data inspection, however, remained a neglected area, taking hours or even days to finalize for particularly complex peptide projects (Figure 1).
Figure 1. Byosphere automation targets the project creation and reporting steps, while data inspection remains the bottleneck.
To reduce data inspection time and increase consistency, a new product was developed: Auto-inspector.
Auto-inspector is designed to reduce data inspection time. Editable fields such as comments, labels and validation fields can be populated automatically, following any custom logic such as a set of rules or a decision tree. Part of this custom logic could involve comparing MS2 spectra against a spectral library populated with historical, curated data.
Here we present an Auto-inspector for PTM analysis, developed to illustrate the underlying principle. The complete logic can be changed entirely to fit a customer's needs; the name "Auto-inspector" is used throughout this article purely for readability, and should not be taken to mean that the product is limited to what is presented here. Other case examples exist for SVA and disulfide analysis and can be shared on request.
PTM Analysis with Auto-inspector
The general steps of Auto-inspector are grouping peptides, filtering within each group, checking chromatographic attributes, comparing MS2 spectra to a spectral library, and finally validating based on peptide attributes.
In the first step, Auto-inspector identifies modified peptides and creates a group named after the protein ID and the protein position of the modification. All other peptides originating from the same protein and containing the same residue are assigned to the same group. Peptides that could be assigned to multiple groups are duplicated across them; however, if multiple groups end up containing the exact same set of peptides, only one of those groups is kept.
Note: The prerequisites for running this particular logic are enabling MS1 correlation during project creation and running "Add missing charge states" and "Add missing via existing" after project creation.
Figure 2. Grouping of peptides.
After grouping, a filter is applied based on the following criteria: intensity, charge state, missed cleavages and, for peptides that exist only as in-silico peptides, MS1 isotope envelope confidence. As long as one peptide with the same sequence (including all modified forms and charge state) exceeds a set threshold in at least one file across the project, all peptides sharing that sequence, modification, and charge state are kept; if none meet the criteria, they are flagged as false positive. This filtering step alone reduced the number of items in the peptide table by 80% on a test project.
Figure 3. Impact of filtering to reduce the number of rows in the peptide table.
Following the filtering step, the remaining peptides are inspected for chromatographic attributes, such as whether a modification arose as an in-source artifact or whether multiple peptides exist with similar XICs. If either is the case, the peptide is flagged as "Uncertain" with a descriptive comment.
MS2 spectra of all remaining peptides are then compared against those stored in a spectral library, which can be populated with curated project files. Peptides whose spectral match exceeds a threshold of 0.9 are set as "True-positive." At the final stage, the remaining peptides are checked against additional peptide attributes such as score and delta mod score. Peptides that fulfill all criteria are set as "True-positive."
Figure 4. Complete decision tree used by Auto-inspector to classify PTM peptides, from initial grouping through final True-positive/Uncertain/False-positive validation.
An auto-inspected file can look like the picture below: the Labels field contains the group name, Comments are automatically populated with a description, and the validation fields are filled in accordingly.
Figure 5. Example peptide table with automatically populated fields.
Customization and Manual Triggering
Thresholds can be customized with written commands that can either be put in the advanced command section during the project creation or in the signature comment.
Manual Triggering
Auto-inspection can be triggered automatically, for example, if the file is created in a certain folder or its name contains a unique marker. Project files can also be auto-inspected by a manual trigger; currently, a change in signature state can trigger inspection of the file.
Signing must be enabled in the Byosphere folder, and a report must be attached to the project file.
Click the edit button in the Byosphere web client (Browse section) for the project file
Click the signature button that appears in the top right of the edit window*
Select the appropriate action, ad comments (optional) and press sign.
A new version of the project file then appears in the file history with the comment "New version uploaded via script."
* The signature button will only appear if signatures have been enabled
Figure 6. Manual Triggering of Auto-inspector.
Conclusion
Auto-inspector demonstrates that data inspection – previously the most time-consuming and least consistent step in a Byosphere project – can be substantially automated. For PTM analysis, a customizable grouping, filtering, and attribute-based decision tree reduced the peptide table by 80% on a test project, while automatically and consistently populating labels, comments, and validation fields. Because the decision logic is fully configurable per customer and can be triggered automatically or via a signature workflow, Auto-inspector can be adapted well beyond PTM analysis; equivalent logic has already been implemented for SVA and disulfide-linked peptide analysis, with further case examples available on request.