Pairwise agonist scanning predicts cellular signaling responses to combinatorial stimuli

Abstract

Patient-specific prediction of cellular response to multiple stimuli is central to evaluating clinical risk, disease progression, and response to therapy. We deployed Pairwise Agonist Scanning (PAS) to measure calcium signaling of human platelets in EDTA-treated plasma exposed to 6 different agonists (at 0.1, 1, and 10×EC50) used individually or in 135 pairwise combinations. With 154 traces, we trained a neural network (NN) model to accurately predict the entire 6-dimensional response to ADP, convulxin, U46619, SFLLRN, AYPGKF, and PGE2. The NN successfully predicted calcium responses to sequential agonist additions, all ternary combinations of [ADP]+[convulxin]+[SFLLRN] (R=0.88), and 45 different combinations of 4 to 6 agonists (R=0.88). Furthermore, PAS provided 135 pairwise synergy values that allowed a unique phenotypic scoring and differentiation of 10 donors. Training of NNs with pairs of stimuli across the dose-response regime represents a highly efficient approach to predict integration of multiple, complex signals in a patient-specific disease milieu.