DNA-methyltransferase inhibitors (DNMTis), such as azacitidine and decitabine, are used clinically to treat myelodysplastic syndrome (MDS) and acute myeloid leukemia (AML). Decitabine activates the transcription of endogenous retroviruses (ERVs), which can induce immune response by acting as cellular double-stranded RNAs (dsRNAs). Yet, the posttranscriptional regulation of ERV dsRNAs remains uninvestigated. Here, we find that the viral mimicry and subsequent cell death in response to decitabine require the dsRNA-binding protein Staufen1 (Stau1). We show that Stau1 directly binds to ERV RNAs and stabilizes them in a genome-wide manner. Furthermore, Stau1-mediated stabilization requires a long noncoding RNA TINCR, which enhances the interaction between Stau1 and ERV RNAs. Analysis of a clinical patient cohort reveals that MDS and AML patients with lower Stau1 and TINCR expressions exhibit inferior treatment outcomes to DNMTi therapy. Overall, our study reveals the posttranscriptional regulatory mechanism of ERVs and identifies the Stau1-TINCR complex as a potential target for predicting the efficacy of DNMTis and other drugs that rely on dsRNAs.

BACKGROUND Substantial increase in the use of Electronic Health Records (EHRs) has opened new frontiers for predictive healthcare. However, while EHR systems are nearly ubiquitous, they lack a unified code system for representing medical concepts. Heterogeneous formats of EHR present a substantial barrier for the training and deployment of state-of-the-art deep learning models at scale. OBJECTIVE The aim of this study is to suggest a novel text embedding approach to overcome heterogeneity of EHR structure among different EHR systems. METHODS We introduce Description-based Embedding, DescEmb, a code-agnostic description-based representation learning framework for predictive modeling on EHR. DescEmb takes advantage of the flexibility of neural language understanding models while maintaining a neutral approach that can be combined with prior frameworks for task-specific representation learning or predictive modeling. RESULTS Based on five prediction tasks with two heterogeneous EHR datasets, DescEmb achieves comparable or superior performance to the traditional code-based embedding approach, especially under the zero-shot and few-shot transfer learning scenarios. We also demonstrate that DescEmb enables us to train a single model on a pooled dataset from heterogeneous EHR systems and achieve the same, if not better performance compared to training separate models for each EHR system. CONCLUSIONS Based on the promising results, we believe the description-based embedding approach on EHR will open a new direction for large-scale predictive modeling in healthcare.

High-intensity aerobic exercise (90% of the maximal heart rate) can effectively suppress cancer cell proliferation in vivo. However, the molecular effects of exercise and its relevance to cancer prevention remain uninvestigated. In this study, mice with colorectal cancer were subjected to high-intensity aerobic exercise, and mRNA-seq analysis was performed on the heart, lungs, and skeletal muscle tissues to analyze the genome-wide molecular effects of exercise. The skeletal muscle-derived genes with exercise-dependent differential expression were further evaluated for their effects on colorectal cancer cell viability. Compared to the results obtained for the control groups (healthy and cancer with no exercise), the regular and high-intensity aerobic physical activity in the mice produced positive results in comprehensive parameters (i.e., food intake, weight gain, and survival rate). A heatmap of differentially expressed genes revealed markedly different gene expression patterns among the groups. RNA-seq analysis of 23,282 genes expressed in the skeletal muscle yielded several anticancer effector genes (e.g., Trim63, Fos, Col1a1, and Six2). Knockdown and overexpression of selected anticancer genes repressed CT26 murine colorectal carcinoma cell proliferation by 20% (p < 0.05). Our findings, based on the aerobic exercise cancer mouse model, suggest that high-intensity aerobic exercise results in a comprehensive change in the expression patterns of genes, particularly those that can affect cancer cell viability. Such an approach may identify key exercise-regulated genes that can help the body combat cancer.

EHR systems lack a unified code system for representing medical concepts, which acts as a barrier for the deployment of deep learning models in large scale to multiple clinics and hospitals. To overcome this problem, we introduce Description-based Embedding, DescEmb, a codeagnostic representation learning framework for EHR. DescEmb takes advantage of the flexibility of neural language understanding models to embed clinical events using their textual descriptions rather than directly mapping each event to a dedicated embedding. DescEmb outperformed traditional code-based embedding in extensive experiments, especially in a zero-shot transfer task (one hospital to another), and was able to train a single unified model for heterogeneous EHR datasets.