Deep learning models for image and data processes of intracellular calcium ions

Abstract

Intracellular calcium ion (Ca2+) in cytoplasm as an intracellular second messenger is involved in almost all important cellular activities of organisms. Generally its concentration ([Ca2+]i) is tested by live imaging followed image and data processes, in which much tedious and subjective manual work is involved. Here we show a computational approach of Deep Calcium following the principles of deep learning to predict the cytoplasmic Ca2+ ranges and calcium peaks in calcium curve of objective cells. To validate Deep Calcium, chondrocytes, bone marrow stromal cells (BMSCs) and osteoblastic like cells (MC3T3-E1) from both the tissue and cell samples as well as from spontaneous and mechanical stimulated calcium response patterns are used. The good performance comparing with other relative machine learning models, as well as consistency biological results with human experts are demonstrated. Deep Calcium provides references for other image and data processes of intracellular range determination and curve peak identification.