CPCB and Bio Ph.D. students co-first author paper accepted in Bioinformatics
CPCB 2nd year student Huangqingbo (Paul) Sun and Biological Sciences Ph.D. student Xuecong Fu are co-first authors on the Murphy group paper "Improving and evaluating deep learning models of cellular organization" that has been accepted by Bioinformatics. The work builds on the innovative deep learning approach to generating organelle patterns from brightfield images developed by former CPCB student Greg Johnson and his collaborators (Greg is currently Head of Machine Learning at NewLimit). The new paper begins by describing novel metrics for evaluating synthetic cell images that can capture the high-level properties of cell organelle patterns to go beyond the pixel-wise measures used in training deep learning models. It then describes ways to improve deep learning generative models of cell organization and introduces an alternative modeling approach that gave the best performance on the new metrics. All results and source code are available as open source.