DETECTION OF REPLICATION FORKS IN EM IMAGES USING FASTER R-CNN

Currently, one of the most effective techniques to study the mechanisms of DNA replication is using the electron microscopy. Typical imaging experiments result in terabytes of image data, where finding and classifying the replication forks is done manually by experts and is extremely time consuming. Here, we present a fully automated deep learning based approach for detection and classification of DNA replication forks. It is based on the Faster R-CNN architecture, following with additional classification layers. The experimental results using real data indicate that the proposed method achieves detection accuracy which approaches the performance of expert annotators, even with limited amounts of training data.