Knowledge-guided fuzzy logic modeling to infer cellular signaling networks from proteomic data

Area of Science:

• Systems Biology
• Computational Biology
• Pharmacology

Background:

• Canonical signaling pathways lack cell-type specificity, limiting predictions of drug responses.
• Purely data-driven network inference methods often lack biological interpretability.
• Hybrid approaches integrating prior knowledge and real data are needed for robust network inference.

Purpose of the Study:

• To develop a knowledge-guided fuzzy logic network model for inferring context-specific signaling pathways.
• To improve the prediction of cellular responses to drug treatments using time-series data.
• To enable precise predictions of anticancer drug effects for precision medicine.

Main Methods:

• Proposed a knowledge-guided fuzzy logic network model integrating prior biological knowledge and time-series data.
• Utilized dynamic time warping to measure the fit between experimental and predicted data for time-series modeling.
• Evaluated the model on synthetic and real phosphoproteomic datasets.

Main Results:

• The model successfully uncovered drug-induced alterations in signaling pathways in cancer cells.
• Demonstrated the ability to model feedback loops within signaling networks from time-series data.
• Showcased improved prediction of context-specific anticancer drug effects compared to existing hybrid models.

Conclusions:

• The developed model offers a powerful tool for understanding and predicting context-specific cellular responses to drugs.
• Calibrating generic signaling pathway models with real data advances precision medicine by enabling accurate drug effect predictions.
• The method facilitates uncovering dynamical mechanisms of signaling networks from time-series experimental observations.