ProteoAutoNet: high-throughput co-eluted protein analysis with robotics and machine learning

Area of Science:

• Biochemistry
• Proteomics
• Computational Biology

Background:

• Co-fractionation mass spectrometry (CF-MS) is crucial for mapping protein-protein interactions.
• Current CF-MS methods suffer from low throughput, limiting predictive model training data.
• Scarcity and limited diversity of high-quality data hinder the development of robust predictive models for protein interactions.

Purpose of the Study:

• To develop ProteoAutoNet, a robotic platform and computational workflow for high-throughput CF-MS analysis.
• To enhance the throughput and data quality of protein-protein interaction studies.
• To improve the robustness and predictive power of machine learning models for protein interactions.

Main Methods:

• Implemented a robotic experimental platform integrated with a computational workflow for CF-MS.
• Utilized targeted data augmentation within a machine learning model to expand and diversify protein interaction data.
• Applied the ProteoAutoNet system to analyze three thyroid cell lines.

Main Results:

• Achieved a two-fold increase in sample processing throughput from protein complex to peptide.
• Predicted 25,173 co-eluted proteins with an area under the receiver operating characteristic curve (AUROC) of 0.78.
• Identified significantly upregulated proteasome and prefoldin complexes in a metastatic thyroid cancer cell line and a novel interaction between TGM2 and HK1.

Conclusions:

• ProteoAutoNet offers an improved framework for high-throughput protein-protein interaction investigation.
• The platform enhances the discovery of biologically relevant protein interactions.
• This approach advances the understanding of cellular mechanisms and disease-associated changes in protein interaction networks.