RESUMEN
Motivation: To predict drug targets, graph-based machine-learning methods have been widely used to capture the relationships between drug, target and disease entities in drug-disease-target (DDT) networks. However, many methods cannot explicitly consider disease types at inference time and so will predict the same target for a given drug under any disease condition. Meanwhile, DDT networks are usually organized hierarchically carrying interactive relationships between involved entities, but these methods, especially those based on Euclidean embedding cannot fully utilize such topological information, which might lead to sub-optimal results. We hypothesized that, by importing hyperbolic embedding specifically for modeling hierarchical DDT networks, graph-based algorithms could better capture relationships between aforementioned entities, which ultimately improves target prediction performance. Results: We formulated the target prediction problem as a knowledge graph completion task explicitly considering disease types. We proposed FLONE, a hyperbolic embedding-based method based on capturing hierarchical topological information in DDT networks. The experimental results on two DDT networks showed that by introducing hyperbolic space, FLONE generates more accurate target predictions than its Euclidean counterparts, which supports our hypothesis. We also devised hyperbolic encoders to fuse external domain knowledge, to make FLONE enable handling samples corresponding to previously unseen drugs and targets for more practical scenarios. Availability and implementation: Source code and dataset information are at: https://github.com/arantir123/DDT_triple_prediction. Supplementary information: Supplementary data are available at Bioinformatics Advances online.
RESUMEN
Drug combinations could trigger pharmacological therapeutic effects (TEs) and adverse effects (AEs). Many computational methods have been developed to predict TEs, e.g. the therapeutic synergy scores of anti-cancer drug combinations, or AEs from drug-drug interactions. However, most of the methods treated the AEs and TEs predictions as two separate tasks, ignoring the potential mechanistic commonalities shared between them. Based on previous clinical observations, we hypothesized that by learning the shared mechanistic commonalities between AEs and TEs, we could learn the underlying MoAs (mechanisms of actions) and ultimately improve the accuracy of TE predictions. To test our hypothesis, we formulated the TE prediction problem as a multi-task heterogeneous network learning problem that performed TE and AE learning tasks simultaneously. To solve this problem, we proposed Muthene (multi-task heterogeneous network embedding) and evaluated it on our collected drug-drug interaction dataset with both TEs and AEs indications. Our experimental results showed that, by including the AE prediction as an auxiliary task, Muthene generated more accurate TE predictions than standard single-task learning methods, which supports our hypothesis. Using a drug pair Vincristine-Dasatinib as a case study, we demonstrated that our method not only provides a novel way of TE predictions but also helps us gain a deeper understanding of the MoAs of drug combinations.
Asunto(s)
Efectos Colaterales y Reacciones Adversas Relacionados con Medicamentos , Humanos , Interacciones Farmacológicas , Combinación de Medicamentos , Aprendizaje AutomáticoRESUMEN
BACKGROUND: To study the association between dynamic iris change and primary angle-closure disease (PACD) with anterior segment optical coherence tomography (AS-OCT) videos and develop an automated deep learning system for angle-closure screening as well as validate its performance. METHODS: A total of 369 AS-OCT videos (19,940 frames)-159 angle-closure subjects and 210 normal controls (two datasets using different AS-OCT capturing devices)-were included. The correlation between iris changes (pupil constriction) and PACD was analyzed based on dynamic clinical parameters (pupil diameter) under the guidance of a senior ophthalmologist. A temporal network was then developed to learn discriminative temporal features from the videos. The datasets were randomly split into training, and test sets and fivefold stratified cross-validation were used to evaluate the performance. RESULTS: For dynamic clinical parameter evaluation, the mean velocity of pupil constriction (VPC) was significantly lower in angle-closure eyes (0.470 mm/s) than in normal eyes (0.571 mm/s) (P < 0.001), as was the acceleration of pupil constriction (APC, 3.512 mm/s2 vs. 5.256 mm/s2; P < 0.001). For our temporal network, the areas under the curve of the system using AS-OCT images, original AS-OCT videos, and aligned AS-OCT videos were 0.766 (95% CI: 0.610-0.923) vs. 0.820 (95% CI: 0.680-0.961) vs. 0.905 (95% CI: 0.802-1.000) (for Casia dataset) and 0.767 (95% CI: 0.620-0.914) vs. 0.837 (95% CI: 0.713-0.961) vs. 0.919 (95% CI: 0.831-1.000) (for Zeiss dataset). CONCLUSIONS: The results showed, comparatively, that the iris of angle-closure eyes stretches less in response to illumination than in normal eyes. Furthermore, the dynamic feature of iris motion could assist in angle-closure classification.
RESUMEN
Semantic segmentation of surgery scenarios is a fundamental task for computer-aided surgery systems. Precise segmentation of surgical instruments and anatomies contributes to capturing accurate spatial information for tracking. However, uneven reflection and class imbalance lead the segmentation in cataract surgery to a challenging task. To desirably conduct segmentation, a network with multi-view decoders (MVD-Net) is proposed to present a generalizable segmentation for cataract surgery. Two discrepant decoders are implemented to achieve multi-view learning with the backbone of U-Net. The experiment is carried out on the Cataract Dataset for Image Segmentation (CaDIS). The ablation study verifies the effectiveness of the proposed modules in MVD-Net, and superior performance is provided by MVD-Net in the comparison with the state-of-the-art methods. The source code will be publicly released.
Asunto(s)
Catarata , Aprendizaje Profundo , Humanos , Procesamiento de Imagen Asistido por Computador/métodos , Redes Neurales de la Computación , SemánticaRESUMEN
Quality degradation (QD) is common in the fundus images collected from the clinical environment. Although diagnosis models based on convolutional neural networks (CNN) have been extensively used to interpret retinal fundus images, their performances under QD have not been assessed. To understand the effects of QD on the performance of CNN-based diagnosis model, a systematical study is proposed in this paper. In our study, the QD of fundus images is controlled by independently or simultaneously importing quantified interferences (e.g., image blurring, retinal artifacts, and light transmission disturbance). And the effects of diabetic retinopathy (DR) grading systems are thus analyzed according to the diagnosis performances on the degraded images. With images degraded by quantified interferences, several CNN-based DR grading models (e.g., AlexNet, SqueezeNet, VGG, DenseNet, and ResNet) are evaluated. The experiments demonstrate that image blurring causes a significant decrease in performance, while the impacts from light transmission disturbance and retinal artifacts are relatively slight. Superior performances are achieved by VGG, DenseNet, and ResNet in the absence of image degradation, and their robustness is presented under the controlled degradation.