Time-Resolved Phosphoproteomics-Guided BFS Beam Search Reveals Cell-Type-Specific EGFR Signaling Architectures and SHP2 Inhibitor-Induced Pathway Rewiring

This study introduces a systematic computational framework that integrates time-resolved phosphoproteomics with a BFS-guided Beam Search algorithm to reconstruct cell-type-specific EGFR signaling networks, successfully revealing how SHP2 inhibition rewires pathway architectures and drives adaptive resistance mechanisms.

The Gist

Imagine your cell as a bustling, high-tech city. Inside this city, there are millions of tiny messengers (proteins) running around, passing notes to tell the city how to react to things like growth signals or drugs. When a drug tries to stop a specific bad actor in the city, the messengers often get clever: they reroute their messages through different streets to keep the city running anyway. This is what scientists call "adaptive resistance."

The problem is that we have a massive amount of data about these messengers (phosphoproteomics), but it's like having a million scattered postcards without a map. We know what was sent, but we don't know exactly which streets the messages took to get there.

The New "GPS" for Cell Cities
This paper introduces a new computer program that acts like a smart GPS to figure out exactly which routes these messages are taking. Here is how it works, using simple analogies:

• The Map (STRING Database): Instead of building a giant library of every possible road in the world, the program connects to a live, online map (the STRING database) that already knows which proteins are friends with each other. It doesn't need to store the whole map on its own hard drive; it just looks up the connections in real-time.
• The Search Strategy (BFS + Beam Search): Imagine you are trying to find a path from the city gate (the start of a signal) to the mayor's office (the final effect).
  • The program uses a Breadth-First Search (BFS) to look at all possible roads at the same time, like a drone scanning every street in a neighborhood simultaneously.
  • However, looking at every possible road in the whole city would take forever. So, it uses a Beam Search. Think of this as a flashlight that only shines on the top 5 most promising roads at any given moment, ignoring the dead ends. It keeps the search focused and fast.
• Filtering the Noise (MAD and Cleaning): Not every note found on the street is important. The program uses a statistical filter (MAD) to decide which notes are real signals and which are just background noise. After it finds all the possible routes, it runs a "cleaning crew" to remove loops (where a message goes in circles) and checks a local directory (Human Protein Atlas) to make sure the buildings on the route actually exist in that specific type of cell.

What They Discovered
The researchers tested this GPS on three different types of "cities" (HeLa, MDA-MB-468, and HEK293T cells). They found that every city has a unique layout; what works in one doesn't work in another.

They specifically looked at what happens when they tried to block a specific traffic cop named SHP2 in the MDA-MB-468 city:

1. The Blockade: When they stopped SHP2, the old main road (PTPN11) was closed.
2. The Detour: The messengers didn't stop; they immediately found new shortcuts. They started using ERBB3 and PIK3CA as their new main entry points much more often.
3. The Recovery: When they removed the drug (washed it out), the city slowly started rebuilding the old SHP2 road, and the traffic shifted back to the original main entry point, ERBB2.

The Bottom Line
This paper doesn't just say "drugs fail." It provides a systematic, reproducible way to draw a detailed map of how cells reroute their signals when under attack. By turning messy data into clear, step-by-step roadmaps, this tool helps scientists understand exactly how cells are outsmarting treatments, which is the first step in designing better strategies to stop them.

Technical Summary

Technical Summary: Time-Resolved Phosphoproteomics-Guided BFS Beam Search

Problem Statement
Adaptive resistance to kinase- and phosphatase-targeted therapies is often driven by the pharmacological rewiring of intracellular signaling networks. However, the field currently lacks systematic computational methods capable of quantifying these cross-condition pathway changes directly from time-resolved phosphoproteomic data. Existing approaches struggle to transform high-dimensional, dynamic phosphoproteomic measurements into mechanistically interpretable, cell-type-specific signaling hypotheses.

Methodology
The authors present an algorithmic framework designed to reconstruct cell-type-specific signaling pathways from time-resolved phosphoproteomic data. The core of this framework integrates several distinct computational strategies:

• Data Preprocessing: The system employs a data-adaptive binary-state assignment based on the Median Absolute Deviation (MAD) to categorize phosphosite activation states.
• Pathway Reconstruction Algorithm: The framework utilizes a Breadth-First Search (BFS) combined with an interaction-weight-guided Beam Search. This traversal is anchored to experimentally supported "active nodes" identified at zone boundaries and terminals within the data.
• Network Integration: The search operates over the STRING protein-protein interaction database (accessed via REST API v12.5), allowing for the inclusion of STRING-inferred bridge proteins as intermediate connectors without requiring local database installation.
• Validation and Cleaning: A post-enumeration path cleaning pipeline ensures the generation of biologically interpretable, acyclic signaling routes. Crucially, the framework performs edge-level validation against Human Protein Atlas-based cell-line expression data to ensure biological relevance.

Key Results
The framework was benchmarked across five published phosphoproteomic datasets spanning three cell types: HeLa, MDA-MB-468, and EGFR Flp-In HEK293T. The analysis demonstrated the following specific findings:

• Cell-Type Specificity: The method successfully captured distinct EGFR signaling architectures unique to each cell type.
• SHP2 Inhibition Effects: In MDA-MB-468 cells subjected to three pharmacological conditions, SHP2 inhibition was shown to abolish PTPN11-mediated pathways. This inhibition induced a shift in the first-hop effector distribution, increasing engagement with ERBB3 (rising from 21.5% to 25.2% of paths) and PIK3CA (rising from 9.2% to 14.3% of paths).
• Reversibility and Recovery: Upon washout of the SHP2 inhibitor, the system revealed a partial recovery of PTPN11-mediated pathways. Concurrently, ERBB2 re-emerged as the dominant first-hop effector, accounting for 30.3% of the paths.

Significance and Claims
The paper positions this framework as a systematic and reproducible approach for converting time-resolved phosphoproteomic measurements into mechanistically interpretable signaling hypotheses. The authors claim direct applicability of this method to drug resistance modeling and the design of combination therapies. By quantifying drug-induced pathway rewiring and identifying specific shifts in effector distribution, the framework aims to bridge the gap between raw phosphoproteomic data and actionable insights into adaptive resistance mechanisms.