Dendral and meta-dendral: Their applications dimension

In order to effectively integrate medical data and alleviate the problem of uneven distribution of medical resources. In this paper, we combine the techniques of expert systems, graph neural networks, and knowledge graphs to propose a disease guidance model combining semi-supervised graph neural networks and knowledge graphs. We use the MASR speech recognition module combined with gated convolutional units for effective text processing of different types of speech; then we use the LTP module in natural language processing for semantic analysis and segmentation matching of interrogative sentences; we combine keywords with the number of diseases and divide and construct the set of nodes with knowledge graphs. And we use semi-supervised graph neural network type analysis to give treatment results and rehabilitation suggestions effectively. We optimize the Chinese and English corpora respectively, adding consideration for local dialect audiences. We performed a comprehensive comparison of the accuracy and training time of several mainstream GCN algorithms and our GCN semi-supervised (SGS) under various graphical text datasets to validate the efficiency and accuracy of our own algorithm choices. We preprocess the number of different symptoms for classification and simplify the redundant nodes to optimize the running time while taking into account the overall convergence. The operational mechanism of the model as well as the convergence and hits under different symptom parameters are explained through hit rate and convergence rate metrics to demonstrate the effectiveness and stability of the model under proprietary medical conditions.

Metabolomics is a powerful tool to rationally guide the metabolic engineering of synthetic bioproduction pathways. Current reports indicate great potential to further develop metabolomics-directed synthetic bioproduction. Advanced mass metabolomics methods including isotope flux analysis, untargeted metabolomics, and system-wide approaches are assisting the characterization of metabolic pathways and enabling the biosynthesis of more complex products. More importantly, a design, build, test, and learn (DBTL) cycle is accelerating synthetic biology research and is highly compatible with metabolomics data to further expand bioproduction capability. However, learning processes are currently the weakest link in this workflow. Therefore, guidelines for the development of metabolic learning processes are proposed based on bioproduction examples. Linking dynamic mass spectrometry (MS) methodologies together with automated learning workflows is encouraged.