Differentiation of Active Corneal Infections from Healed Scars Using Deep Learning.

PURPOSE: To develop and evaluate an automated, portable algorithm to differentiate active corneal ulcers from healed scars using only external photographs. DESIGN: A convolutional neural network was trained and tested using photographs of corneal ulcers and scars. PARTICIPANTS: De-identified photographs of corneal ulcers were obtained from the Steroids for Corneal Ulcers Trial (SCUT), Mycotic Ulcer Treatment Trial (MUTT), and Byers Eye Institute at Stanford University. METHODS: Photographs of corneal ulcers (n = 1313) and scars (n = 1132) from the SCUT and MUTT were used to train a convolutional neural network (CNN). The CNN was tested on 2 different patient populations from eye clinics in India (n = 200) and the Byers Eye Institute at Stanford University (n = 101). Accuracy was evaluated against gold standard clinical classifications. Feature importances for the trained model were visualized using gradient-weighted class activation mapping. MAIN OUTCOME MEASURES: Accuracy of the CNN was assessed via F1 score. The area under the receiver operating characteristic (ROC) curve (AUC) was used to measure the precision-recall trade-off. RESULTS: The CNN correctly classified 115 of 123 active ulcers and 65 of 77 scars in patients with corneal ulcer from India (F1 score, 92.0% [95% confidence interval (CI), 88.2%-95.8%]; sensitivity, 93.5% [95% CI, 89.1%-97.9%]; specificity, 84.42% [95% CI, 79.42%-89.42%]; ROC: AUC, 0.9731). The CNN correctly classified 43 of 55 active ulcers and 42 of 46 scars in patients with corneal ulcers from Northern California (F1 score, 84.3% [95% CI, 77.2%-91.4%]; sensitivity, 78.2% [95% CI, 67.3%-89.1%]; specificity, 91.3% [95% CI, 85.8%-96.8%]; ROC: AUC, 0.9474). The CNN visualizations correlated with clinically relevant features such as corneal infiltrate, hypopyon, and conjunctival injection. CONCLUSIONS: The CNN classified corneal ulcers and scars with high accuracy and generalized to patient populations outside of its training data. The CNN focused on clinically relevant features when it made a diagnosis. The CNN demonstrated potential as an inexpensive diagnostic approach that may aid triage in communities with limited access to eye care.

The ATM- and ATR-related SCD domain is over-represented in proteins involved in nervous system development.

ATM and ATR are cellular kinases with a well-characterized role in the DNA-damage response. Although the complete set of ATM/ATR targets is unknown, they often contain clusters of S/TQ motifs that constitute an SCD domain. In this study, we identified putative ATM/ATR targets that have a conserved SCD domain across vertebrates. Using this approach, we have identified novel putative ATM/ATR targets in pathways known to be under direct control of these kinases. Our analysis has also unveiled significant enrichment of SCD-containing proteins in cellular pathways, such as vesicle trafficking and actin cytoskeleton, where a regulating role for ATM/ATR is either unknown or poorly understood, hinting at a much broader and overarching role for these kinases in the cell. Of particular note is the overrepresentation of conserved SCD-containing proteins involved in pathways related to neural development. This finding suggests that ATM/ATR could be directly involved in controlling this process, which may be linked to the adverse neurological effects observed in patients with mutations in ATM.

SCDFinder, a web-based tool for the identification of putative novel ATM and ATR targets.

MOTIVATION: The S/TQ cluster domain (SCD) constitutes a new type of protein domain that is not defined by sequence similarity but by the presence of multiple S/TQ motifs within a variable stretch of amino acids. SCDs are recognized targets for DNA damage response (DDR) kinases like ATM and ATR. Characterizing DDR targets is of significant interest. The aim of this work was to develop a web-based tool to allow for easy identification and visualization of SCDs within specific proteins or in whole proteome sets, a feature not supported by current domain and motif search tools. RESULTS: We have developed an algorithm that (i) generates a list of all proteins in an organism containing at least one user-defined SCD within their sequence, or (ii) identifies and renders a visual representation of all user-defined SCDs present in a single sequence or batch of sequences. AVAILABILITY AND IMPLEMENTATION: The application was developed using Pearl and Python, and is available at the following URL: http://ustbioinfo.webfactional.com/scd/.