Extended Modelling of Molecular Calcium Signalling in Platelets by Combined Recurrent Neural Network and Partial...

Area of Science:

• Biochemistry
• Computational Biology
• Hematology

Background:

• Platelets are crucial for hemostasis and thrombosis.
• Agonist-induced cytosolic calcium ([Ca2+]i) rises drive platelet activation.
• The precise patterns of platelet calcium responses remain incompletely understood.

Purpose of the Study:

• To develop and apply computational models for analyzing platelet calcium ([Ca2+]i) dynamics.
• To accurately predict platelet calcium responses under various stimulation and inhibition conditions.
• To identify key variables influencing platelet calcium signaling for potential therapeutic targets.

Main Methods:

• Utilized Fura-2 loaded platelets stimulated with agonists (thrombin, collagen, CRP).
• Developed and trained non-linear models: Multilayer Perceptron (MLP) and NARX (autoregressive network with exogenous inputs).
• Constructed a Partial Least Square (PLS) linear regression model for variable importance estimation.

Main Results:

• Trained networks accurately predicted [Ca2+]i curves for agonist and inhibitor combinations.
• The NARX model achieved an R2 of 0.64 for predicting unforeseen data trends.
• The PLS model provided interpretable insights into the importance of individual variables.

Conclusions:

• Computational modeling, particularly NARX, effectively captures nanomolar [Ca2+]i curves in platelets.
• PLS modeling offers valuable, interpretable data on platelet signaling pathways.
• These models can inform the development of novel platelet [Ca2+]i-inhibiting drugs and aid in diagnosing bleeding disorders.