Exploiting biochemical data to improve osteosarcoma diagnosis with deep learning.

Early and accurate diagnosis of osteosarcomas (OS) is of great clinical significance, and machine learning (ML) based methods are increasingly adopted. However, current ML-based methods for osteosarcoma diagnosis consider only X-ray images, usually fail to generalize to new cases, and lack explainability. In this paper, we seek to explore the capability of deep learning models in diagnosing primary OS, with higher accuracy, explainability, and generality. Concretely, we analyze the added value of integrating the biochemical data, i.e., alkaline phosphatase (ALP) and lactate dehydrogenase (LDH), and design a model that incorporates the numerical features of ALP and LDH and the visual features of X-ray imaging through a late fusion approach in the feature space. We evaluate this model on real-world clinic data with 848 patients aged from 4 to 81. The experimental results reveal the effectiveness of incorporating ALP and LDH simultaneously in a late fusion approach, with the accuracy of the considered 2608 cases increased to 97.17%, compared to 94.35% in the baseline. Grad-CAM visualizations consistent with orthopedic specialists further justified the model's explainability.

Cascade and Fusion: A Deep Learning Approach for Camouflaged Object Sensing.

The demand for the sensor-based detection of camouflage objects widely exists in biological research, remote sensing, and military applications. However, the performance of traditional object detection algorithms is limited, as they are incapable of extracting informative parts from low signal-to-noise ratio features. To address this problem, we propose Camouflaged Object Detection with Cascade and Feedback Fusion (CODCEF), a deep learning framework based on an RGB optical sensor that leverages a cascaded structure with Feedback Partial Decoders (FPD) instead of a traditional encoder-decoder structure. Through a selective fusion strategy and feedback loop, FPD reduces the loss of information and the interference of noises in the process of feature interweaving. Furthermore, we introduce Pixel Perception Fusion (PPF) loss, which aims to pay more attention to local pixels that might become the edges of an object. Experimental results on an edge device show that CODCEF achieved competitive results compared with 10 state-of-the-art methods.

Drug-Drug Interaction Predictions via Knowledge Graph and Text Embedding: Instrument Validation Study.

BACKGROUND: Minimizing adverse reactions caused by drug-drug interactions (DDIs) has always been a prominent research topic in clinical pharmacology. Detecting all possible interactions through clinical studies before a drug is released to the market is a demanding task. The power of big data is opening up new approaches to discovering various DDIs. However, these data contain a huge amount of noise and provide knowledge bases that are far from being complete or used with reliability. Most existing studies focus on predicting binary DDIs between drug pairs and ignore other interactions. OBJECTIVE: Leveraging both drug knowledge graphs and biomedical text is a promising pathway for rich and comprehensive DDI prediction, but it is not without issues. Our proposed model seeks to address the following challenges: data noise and incompleteness, data sparsity, and computational complexity. METHODS: We propose a novel framework, Predicting Rich DDI, to predict DDIs. The framework uses graph embedding to overcome data incompleteness and sparsity issues to make multiple DDI label predictions. First, a large-scale drug knowledge graph is generated from different sources. The knowledge graph is then embedded with comprehensive biomedical text into a common low-dimensional space. Finally, the learned embeddings are used to efficiently compute rich DDI information through a link prediction process. RESULTS: To validate the effectiveness of the proposed framework, extensive experiments were conducted on real-world data sets. The results demonstrate that our model outperforms several state-of-the-art baseline methods in terms of capability and accuracy. CONCLUSIONS: We propose a novel framework, Predicting Rich DDI, to predict DDIs. Using rich DDI information, it can competently predict multiple labels for a pair of drugs across numerous domains, ranging from pharmacological mechanisms to side effects. To the best of our knowledge, this framework is the first to provide a joint translation-based embedding model that learns DDIs by integrating drug knowledge graphs and biomedical text simultaneously in a common low-dimensional space. The model also predicts DDIs using multiple labels rather than single or binary labels. Extensive experiments were conducted on real-world data sets to demonstrate the effectiveness and efficiency of the model. The results show our proposed framework outperforms several state-of-the-art baselines.

DEEP MOTIF DASHBOARD: VISUALIZING AND UNDERSTANDING GENOMIC SEQUENCES USING DEEP NEURAL NETWORKS.

Deep neural network (DNN) models have recently obtained state-of-the-art prediction accuracy for the transcription factor binding (TFBS) site classification task. However, it remains unclear how these approaches identify meaningful DNA sequence signals and give insights as to why TFs bind to certain locations. In this paper, we propose a toolkit called the Deep Motif Dashboard (DeMo Dashboard) which provides a suite of visualization strategies to extract motifs, or sequence patterns from deep neural network models for TFBS classification. We demonstrate how to visualize and understand three important DNN models: convolutional, recurrent, and convolutional-recurrent networks. Our first visualization method is finding a test sequence's saliency map which uses first-order derivatives to describe the importance of each nucleotide in making the final prediction. Second, considering recurrent models make predictions in a temporal manner (from one end of a TFBS sequence to the other), we introduce temporal output scores, indicating the prediction score of a model over time for a sequential input. Lastly, a class-specific visualization strategy finds the optimal input sequence for a given TFBS positive class via stochastic gradient optimization. Our experimental results indicate that a convolutional-recurrent architecture performs the best among the three architectures. The visualization techniques indicate that CNN-RNN makes predictions by modeling both motifs as well as dependencies among them.