Machine Learning-Driven Discovery of Structurally Related Natural Products as Activators of the Cardiac Calcium Pump SERCA2a.

Abstract

A key molecular dysfunction in heart failure is the reduced activity of the cardiac sarcoplasmic reticulum Ca²⁺-ATPase (SERCA2a) in cardiac muscle cells. Reactivating SERCA2a improves cardiac function in heart failure models, making it a validated target and an attractive therapeutic approach for heart failure therapy. However, finding small-molecule SERCA2a activators is challenging. In this study, we used a machine learning-based virtual screening to identify SERCA2a activators among 57,423 natural products. The machine learning model identified ten structurally related natural products from Zingiber officinale, Aframomum melegueta, Alpinia officinarum, Alpinia oxyphylla, and Capsicum (chili peppers) as SERCA2a activators. Initial ATPase assays showed seven of these activate SERCA at low micromolar concentrations. Notably, two natural products, Yakuchinone A and Alpinoid D displayed robust concentration-dependent responses in primary ATPase activity assays, efficient lipid bilayer binding and permeation in atomistic simulations, and enhanced intracellular Ca²⁺ transport in adult mouse cardiac cells. While these natural products exert off-target effects on Ca²⁺ signaling, these compounds offer promising avenues for the design and optimization of lead compounds. In conclusion, this study increases the array of calcium pump effectors and provides new scaffolds for the development of novel SERCA2a activators as new therapies for heart failure.