An adjoint-based method for the computation of gradients in coagulation schemes

An adjoint-based methodology is proposed to compute the gradient of the outcomes of mathematical models for the coagulation cascade. The method is first exposed and validated by considering a simple, analytically tractable case involving only 3 species. Its potential is further illustrated by considering a detailed model for the extrinsic pathway involving 34 chemical species interacting through 45 chemical reactions and for which the gradient of Endogeneous Thrombin Potential, clotting time, maximum rate and peak value of thrombin with respect to the initial concentrations and reactions rates are computed. It is shown that the method produces gradients estimates that are fully consistent with the finite differences approximation used so far in the literature, but at a much lower computational cost.

Automated summary

An adjoint-based methodology is proposed to compute the gradient of outcomes in the coagulation cascade mathematical models. The method is validated using a simple case involving 3 species and further applied to a complex model with 34 species and 45 reactions. The results show that the method produces consistent gradient estimates at a lower computational cost compared to the finite differences approximation.