Stroke mortality prediction based on ensemble learning and the combination of structured and textual data.

For severe cerebrovascular diseases such as stroke, the prediction of short-term mortality of patients has tremendous medical significance. In this study, we combined machine learning models Random Forest classifier (RF), Adaptive Boosting (AdaBoost), Extremely Randomised Trees (ExtraTree) classifier, XGBoost classifier, TabNet, and DistilBERT to construct a multi-level prediction model that used bioassay data and radiology text reports from haemorrhagic and ischaemic stroke patients to predict six-month mortality. The performances of the prediction models were measured using the area under the receiver operating characteristic curve (AUROC), the area under the precision-recall curve (AUPRC), precision, recall, and F1-score. The prediction models were built with the use of data from 19,616 haemorrhagic stroke patients and 50,178 ischaemic stroke patients. Novel six-month mortality prediction models for these patients were developed, which enhanced the performance of the prediction models by combining laboratory test data, structured data, and textual radiology report data. The achieved performances were as follows: AUROC = 0.89, AUPRC = 0.70, precision = 0.52, recall = 0.78, and F1 score = 0.63 for haemorrhagic patients, and AUROC = 0.88, AUPRC = 0.54, precision = 0.34, recall = 0.80, and F1 score = 0.48 for ischaemic patients. Such models could be used for mortality risk assessment and early identification of high-risk stroke patients. This could contribute to more efficient utilisation of healthcare resources for stroke survivors.

Lung adenocarcinoma-related target gene prediction and drug repositioning.

Lung cancer is the leading cause of cancer deaths globally, and lung adenocarcinoma (LUAD) is the most common type of lung cancer. Gene dysregulation plays an essential role in the development of LUAD. Drug repositioning based on associations between drug target genes and LUAD target genes are useful to discover potential new drugs for the treatment of LUAD, while also reducing the monetary and time costs of new drug discovery and development. Here, we developed a pipeline based on machine learning to predict potential LUAD-related target genes through established graph attention networks (GATs). We then predicted potential drugs for the treatment of LUAD through gene coincidence-based and gene network distance-based methods. Using data from 535 LUAD tissue samples and 59 precancerous tissue samples from The Cancer Genome Atlas, 48,597 genes were identified and used for the prediction model building of the GAT. The GAT model achieved good predictive performance, with an area under the receiver operating characteristic curve of 0.90. 1,597 potential LUAD-related genes were identified from the GAT model. These LUAD-related genes were then used for drug repositioning. The gene overlap and network distance with the target genes were calculated for 3,070 drugs and 672 preclinical compounds approved by the US Food and Drug Administration. At which, bromoethylamine was predicted as a novel potential preclinical compound for the treatment of LUAD, and cimetidine and benzbromarone were predicted as potential therapeutic drugs for LUAD. The pipeline established in this study presents new approach for developing targeted therapies for LUAD.