DRRisk: A Web-based tool to Assess the Risk of Diabetic Retinopathy through Machine Learning on Electronic Health Records.

Objective: We developed a web-based tool for diabetic retinopathy (DR) risk assessment called DRRisk (https://drandml.cdrewu.edu/) using machine learning on electronic health record (EHR) data, with a goal of preventing vision loss in persons with diabetes, especially in underserved settings. Methods: DRRisk uses Python's Flask framework. Its user-interface is implemented using HTML, CSS and Javascript. Clinical experts were consulted on the tool's design. Results: DRRisk assesses current DR risk for people with diabetes, categorizing their risk level as low, moderate, or high, depending on the percentage of DR risk assigned by the underlying machine learning model. Discussion: A goal of our tool is to help providers prioritize patients at high risk for DR in order to prevent blindness. Conclusion: Our tool uses DR risk factors from EHR data to calculate a diabetic person's current DR risk. It may be useful for identifying unscreened diabetic patients who have undiagnosed DR.

Detecting diabetic retinopathy through machine learning on electronic health record data from an urban, safety net healthcare system.

OBJECTIVE: Clinical guidelines recommend annual eye examinations to detect diabetic retinopathy (DR) in patients with diabetes. However, timely DR detection remains a problem in medically underserved and under-resourced settings in the United States. Machine learning that identifies patients with latent/undiagnosed DR could help to address this problem. MATERIALS AND METHODS: Using electronic health record data from 40 631 unique diabetic patients seen at Los Angeles County Department of Health Services healthcare facilities between January 1, 2015 and December 31, 2017, we compared ten machine learning environments, including five classifier models, for assessing the presence or absence of DR. We also used data from a distinct set of 9300 diabetic patients seen between January 1, 2018 and December 31, 2018 as an external validation set. RESULTS: Following feature subset selection, the classifier with the best AUC on the external validation set was a deep neural network using majority class undersampling, with an AUC of 0.8, the sensitivity of 72.17%, and specificity of 74.2%. DISCUSSION: A deep neural network produced the best AUCs and sensitivity results on the test set and external validation set. Models are intended to be used to screen guideline noncompliant diabetic patients in an urban safety-net setting. CONCLUSION: Machine learning on diabetic patients' routinely collected clinical data could help clinicians in safety-net settings to identify and target unscreened diabetic patients who potentially have undiagnosed DR.

Predictive Models for Diabetic Retinopathy from Non-Image Teleretinal Screening Data.

Introduction: Timely diabetic retinopathy detection remains a problem in medically underserved settings in the US; diabetic patients in these locales have limited access to eye specialists. Teleretinal screening programs have been introduced to address this problem. Methods: Using data on ethnicity, gender, age, hemoglobin A1C, insulin dependence, time since last eye examination, subjective diabetes control, and years with diabetes from 27,116 diabetic patients participating in a Los Angeles County teleretinal screening program, we compared different machine learning methods for predicting retinopathy. The dataset exhibited a class imbalance. Results: Six classifiers learned on the data were predictive of retinopathy. The best model had an AUC of 0.754, sensitivity of 58% and specificity of 80%. Discussion: Successfully detecting retinopathy from diabetic patients' routinely collected clinical data could help clinicians in medically underserved areas identify unscreened diabetic patients who are at risk of developing retinopathy. This work is a step towards that goal.

Structural Basis of Arp2/3 Complex Inhibition by GMF, Coronin, and Arpin.

The evolutionarily conserved Arp2/3 complex plays a central role in nucleating the branched actin filament arrays that drive cell migration, endocytosis, and other processes. To better understand Arp2/3 complex regulation, we used single-particle electron microscopy to compare the structures of Arp2/3 complex bound to three different inhibitory ligands: glia maturation factor (GMF), Coronin, and Arpin. Although the three inhibitors have distinct binding sites on Arp2/3 complex, they each induced an "open" nucleation-inactive conformation. Coronin promoted a standard (previously described) open conformation of Arp2/3 complex, with the N-terminal ß-propeller domain of Coronin positioned near the p35/ARPC2 subunit of Arp2/3 complex. GMF induced two distinct open conformations of Arp2/3 complex, which correlated with the two suggested binding sites for GMF. Furthermore, GMF synergized with Coronin in inhibiting actin nucleation by Arp2/3 complex. Arpin, which uses VCA-related acidic (A) motifs to interact with the Arp2/3 complex, induced the standard open conformation, and two new masses appeared at positions near Arp2 and Arp3. Furthermore, Arpin showed additive inhibitory effects on Arp2/3 complex with Coronin and GMF. Together, these data suggest that Arp2/3 complex conformation is highly polymorphic and that its activities can be controlled combinatorially by different inhibitory ligands.

GMF severs actin-Arp2/3 complex branch junctions by a cofilin-like mechanism.

BACKGROUND: Branched actin filament networks driving cell motility, endocytosis, and intracellular transport are assembled in seconds by the Arp2/3 complex and must be equally rapidly debranched and turned over. One of the only factors known to promote debranching of actin networks is the yeast homolog of glia maturation factor (GMF), which is structurally related to the actin filament-severing protein cofilin. However, the identity of the molecular mechanism underlying debranching and whether this activity extends to mammalian GMF have remained open questions. RESULTS: Using scanning mutagenesis and total internal reflection fluorescence microscopy, we show that GMF depends on two separate surfaces for debranching. One is analogous to the G-actin and F-actin binding site on cofilin, but we show using fluorescence anisotropy and chemical crosslinking that it instead interacts with actin-related proteins in the Arp2/3 complex. The other is analogous to a second F-actin binding site on cofilin, which in GMF appears to contact the first actin subunit in the daughter filament. We further show that GMF binds to the Arp2/3 complex with low nanomolar affinity and promotes the open conformation. Finally, we show that this debranching activity and mechanism are conserved for mammalian GMF. CONCLUSIONS: GMF debranches filaments by a mechanism related to cofilin-mediated severing, but in which GMF has evolved to target molecular junctions between actin-related proteins in the Arp2/3 complex and actin subunits in the daughter filament of the branch. This activity and mechanism are conserved in GMF homologs from evolutionarily distant species.

Coronin 2A mediates actin-dependent de-repression of inflammatory response genes.

Toll-like receptors (TLRs) function as initiators of inflammation through their ability to sense pathogen-associated molecular patterns and products of tissue damage. Transcriptional activation of many TLR-responsive genes requires an initial de-repression step in which nuclear receptor co-repressor (NCoR) complexes are actively removed from the promoters of target genes to relieve basal repression. Ligand-dependent SUMOylation of liver X receptors (LXRs) has been found to suppress TLR4-induced transcription potently by preventing the NCoR clearance step, but the underlying mechanisms remain enigmatic. Here we provide evidence that coronin 2A (CORO2A), a component of the NCoR complex of previously unknown function, mediates TLR-induced NCoR turnover by a mechanism involving interaction with oligomeric nuclear actin. SUMOylated LXRs block NCoR turnover by binding to a conserved SUMO2/SUMO3-interaction motif in CORO2A and preventing actin recruitment. Intriguingly, the LXR transrepression pathway can itself be inactivated by inflammatory signals that induce calcium/calmodulin-dependent protein kinase IIγ (CaMKIIγ)-dependent phosphorylation of LXRs, leading to their deSUMOylation by the SUMO protease SENP3 and release from CORO2A. These findings uncover a CORO2A-actin-dependent mechanism for the de-repression of inflammatory response genes that can be differentially regulated by phosphorylation and by nuclear receptor signalling pathways that control immunity and homeostasis.

Functional surfaces on the actin-binding protein coronin revealed by systematic mutagenesis.

Coronin is a conserved actin-binding protein that co-functions with ADF/cofilin and Arp2/3 complex to govern cellular actin dynamics. Despite emerging roles for coronin in a range of physiological processes and disease states, a detailed understanding of the molecular interactions of coronin with actin and other binding partners has been lacking. Here, we performed a systematic mutational analysis of surfaces on the yeast coronin ß-propeller domain, which binds to F-actin and is conserved in all coronin family members. We generated 21 mutant alleles and analyzed their biochemical effects on actin binding and ADF/cofilin activity. Conserved actin-binding residues mapped to a discrete ridge stretching across one side of the ß-propeller. Mutants defective in actin binding showed loss of synergy with ADF/cofilin in severing filaments, diminished localization to actin structures in vivo, and loss of coronin overexpression growth defects. In addition, one allele showed normal actin binding but impaired functional interactions with ADF/cofilin. Another allele showed normal actin binding but failed to cause coronin overexpression defects. Together, these results indicate that actin binding is critical for many of the biochemical and cellular functions of coronin and that the ß-propeller domain mediates additional functional interactions with ADF/cofilin and possibly other ligands. Conservation of the actin-binding surfaces across distant species and in all three major classes of coronin isoforms suggests that the nature of the coronin-actin association may be similar in other family members.

GMF is a cofilin homolog that binds Arp2/3 complex to stimulate filament debranching and inhibit actin nucleation.

Cell locomotion and endocytosis are powered by the rapid polymerization and turnover of branched actin filament networks nucleated by Arp2/3 complex. Although a large number of cellular factors have been identified that stimulate Arp2/3 complex-mediated actin nucleation, only a small number of studies so far have addressed which factors promote actin network debranching. Here, we investigated the function of a conserved homolog of ADF/cofilin, glia maturation factor (GMF). We found that S. cerevisiae GMF (also called Aim7) localizes in vivo to cortical actin patches and displays synthetic genetic interactions with ADF/cofilin. However, GMF lacks detectable actin binding or severing activity and instead binds tightly to Arp2/3 complex. Using in vitro evanescent wave microscopy, we demonstrated that GMF potently stimulates debranching of actin filaments produced by Arp2/3 complex. Further, GMF inhibits nucleation of new daughter filaments. Together, these data suggest that GMF binds Arp2/3 complex to both "prune" daughter filaments at the branch points and inhibit new actin assembly. These activities and its genetic interaction with ADF/cofilin support a role for GMF in promoting the remodeling and/or disassembly of branched networks. Therefore, ADF/cofilin and GMF, members of the same superfamily, appear to have evolved to interact with actin and actin-related proteins, respectively, and to make mechanistically distinct contributions to the remodeling of cortical actin structures.

Coronin switches roles in actin disassembly depending on the nucleotide state of actin.

Rapid and polarized turnover of actin networks is essential for motility, endocytosis, cytokinesis, and other cellular processes. However, the mechanisms that provide tight spatiotemporal control of actin disassembly remain poorly understood. Here, we show that yeast coronin (Crn1) makes a unique contribution to this process by differentially interacting with and regulating the effects of cofilin on ATP/ADP+P(i) versus ADP actin filaments. Crn1 potently blocks cofilin severing of newly assembled (ATP/ADP+P(i)) filaments but synergizes with cofilin to sever older (ADP) filaments. Thus, Crn1 has qualitatively distinct/opposite effects on actin dynamics depending on the nucleotide state of actin. This bimodal mechanism requires two separate actin-binding domains in Crn1. Consistent with these activities, Crn1 excludes GFP-Cof1 from newly assembled regions of actin networks in vivo and accelerates cellular actin turnover by four fold. We conclude that coronin polarizes the spatial distribution and activity of cofilin to promote selective disassembly of older actin filaments.

Coronin: the double-edged sword of actin dynamics.

Coronin is a conserved actin binding protein that promotes cellular processes that rely on rapid remodeling of the actin cytoskeleton, including endocytosis and cell motility. However, the exact mechanism by which coronin contributes to actin dynamics has remained elusive for many years. Here, we integrate observations from many groups and propose a unified model to explain how coronin controls actin dynamics through coordinated effects on Arp2/3 complex and cofilin. At the front end of actin networks, coronin protects new (ATP-rich) filaments from premature disassembly by cofilin and recruits Arp2/3 complex to filament sides, leading to nucleation, branching and network expansion. At the rear of networks, coronin has strikingly different activities, synergizing with cofilin to dismantle old (ADP-rich) filaments. Thus, coronin spatially targets Arp2/3 complex and cofilin to opposite ends of actin networks. The net effect of coronin's activities is acceleration of polarized actin subunit flux through filamentous arrays. This increases actin network plasticity and replenishes the actin monomer pool required for new filament growth.

Four novel suppressors of gic1 gic2 and their roles in cytokinesis and polarized cell growth in Saccharomyces cerevisiae.

Gic1 and Gic2 are two Cdc42/Rac interactive binding (CRIB) domain-containing effectors of Cdc42-GTPase that promote polarized cell growth in S. cerevisiae. To identify novel genes that functionally interact with Gic1 and Gic2, we screened for high-copy suppressors of a gic1 gic2 temperature-sensitive strain. We identified two pairs of structurally related genes, SKG6-TOS2 and VHS2-MLF3. These genes have been implicated in polarized cell growth, but their functions have not previously been characterized. We found that overproduction of Skg6 and Tos2 in wild-type cells causes aberrant localization of Cdc3 septin and actin structures as well as defective recruitment of Hof1 and impaired formation of the septum at the mother-bud neck. These data suggest a negative regulatory function for Skg6 and Tos2 in cytokinesis. Consistent with this model, deletion of SKG6 suppresses the growth defects associated with loss of HOF1, a positive regulator of cytokinesis. Our analysis of the second pair of gic1 gic2 suppressors, VHS2 and MLF3, suggests that they regulate polarization of the actin cytoskeleton and cell growth and function in a pathway distinct from and parallel to GIC1 and GIC2.