A systematic assessment of deep learning methods for drug response prediction: from in vitro to clinical applications.

Drug response prediction is an important problem in personalized cancer therapy. Among various newly developed models, significant improvement in prediction performance has been reported using deep learning methods. However, systematic comparisons of deep learning methods, especially of the transferability from preclinical models to clinical cohorts, are currently lacking. To provide a more rigorous assessment, the performance of six representative deep learning methods for drug response prediction using nine evaluation metrics, including the overall prediction accuracy, predictability of each drug, potential associated factors and transferability to clinical cohorts, in multiple application scenarios was benchmarked. Most methods show promising prediction within cell line datasets, and TGSA, with its lower time cost and better performance, is recommended. Although the performance metrics decrease when applying models trained on cell lines to patients, a certain amount of power to distinguish clinical response on some drugs can be maintained using CRDNN and TGSA. With these assessments, we provide a guidance for researchers to choose appropriate methods, as well as insights into future directions for the development of more effective methods in clinical scenarios.

Systematic identification of rabbit LncRNAs reveals functional roles in atherosclerosis.

Long noncoding RNAs (lncRNAs) have been gradually emerging as important regulators in various biological processes and diseases, while the contributions of lncRNAs to atherosclerosis remain largely unknown. Our previous work has discovered atherosclerosis associated protein-coding genes by transcriptome sequencing of rabbit models. Here we investigated the roles of lncRNAs in atherosclerosis. We defined a stringent set of 3736 multi-exonic lncRNA transcripts in rabbits. All lncRNAs are firstly reported and 609 (16.3%) of them are conserved in 13 species. Rabbit lncRNAs have similar characteristics to lncRNAs in other mammals, such as relatively short length, low expression, and highly tissue-specificity. The integrative analysis of lncRNAs and co-expressed genes characterize diverse functions of lncRNAs. Comparing two kinds of atherosclerosis models (LDLR-deficient WHHL rabbits and cholesterol-fed NZW rabbits) with their corresponding controls, we found the expression changes of two rabbit models were similar in aorta in but different in liver. The shared change in aorta revealed a subset of lncRNAs involved in immune response, while the cholesterol-fed NZW rabbits showed broader lncRNA expression changes in skeletal muscle system compared to WHHL rabbits. These atherosclerosis-associated lncRNAs and genes provide hits for the experimental validation of lncRNA functions. In summary, our study systematically identified rabbit lncRNAs for the first time and provides new insights for understanding the functions of lncRNAs in atherosclerosis. This article is part of a Special Issue entitled: Accelerating Precision Medicine through Genetic and Genomic Big Data Analysis edited by Yudong Cai & Tao Huang.

PDXliver: a database of liver cancer patient derived xenograft mouse models.

BACKGROUND: Liver cancer is the second leading cause of cancer-related deaths and characterized by heterogeneity and drug resistance. Patient-derived xenograft (PDX) models have been widely used in cancer research because they reproduce the characteristics of original tumors. However, the current studies of liver cancer PDX mice are scattered and the number of available PDX models are too small to represent the heterogeneity of liver cancer patients. To improve this situation and to complement available PDX models related resources, here we constructed a comprehensive database, PDXliver, to integrate and analyze liver cancer PDX models. DESCRIPTION: Currently, PDXliver contains 116 PDX models from Chinese liver cancer patients, 51 of them were established by the in-house PDX platform and others were curated from the public literatures. These models are annotated with complete information, including clinical characteristics of patients, genome-wide expression profiles, germline variations, somatic mutations and copy number alterations. Analysis of expression subtypes and mutated genes show that PDXliver represents the diversity of human patients. Another feature of PDXliver is storing drug response data of PDX mice, which makes it possible to explore the association between molecular profiles and drug sensitivity. All data can be accessed via the Browse and Search pages. Additionally, two tools are provided to interactively visualize the omics data of selected PDXs or to compare two groups of PDXs. CONCLUSION: As far as we known, PDXliver is the first public database of liver cancer PDX models. We hope that this comprehensive resource will accelerate the utility of PDX models and facilitate liver cancer research. The PDXliver database is freely available online at: http://www.picb.ac.cn/PDXliver/.

Large-scale pharmacogenomic studies and drug response prediction for personalized cancer medicine.

The response rate of most anti-cancer drugs is limited because of the high heterogeneity of cancer and the complex mechanism of drug action. Personalized treatment that stratifies patients into subgroups using molecular biomarkers is promising to improve clinical benefit. With the accumulation of preclinical models and advances in computational approaches of drug response prediction, pharmacogenomics has made great success over the last 20 years and is increasingly used in the clinical practice of personalized cancer medicine. In this article, we first summarize FDA-approved pharmacogenomic biomarkers and large-scale pharmacogenomic studies of preclinical cancer models such as patient-derived cell lines, organoids, and xenografts. Furthermore, we comprehensively review the recent developments of computational methods in drug response prediction, covering network, machine learning, and deep learning technologies and strategies to evaluate immunotherapy response. In the end, we discuss challenges and propose possible solutions for further improvement.

Population-based study of effectiveness of neoadjuvant radiotherapy on survival in US rectal cancer patients according to age.

Recent cancer researches pay more attention to younger patients due to the variable treatment response among different age groups. Here we investigated the effectiveness of neoadjuvant radiation on the survival of younger and older patients in stage II/III rectal cancer. Data was obtained from Surveillance, Epidemiology, and End Results (SEER) database (n = 12801). Propensity score matching was used to balance baseline covariates according to the status of neoadjuvant radiation. Our results showed that neoadjuvant radiation had better survival benefit (Log-rank P = 3.25e-06) and improved cancer-specific 3-year (87.6%; 95% CI: 86.4-88.7% vs. 84.1%; 95% CI: 82.8-85.3%) and 5-year survival rates (78.1%; 95% CI: 76.2-80.1% vs. 77%; 95% CI: 75.3-78.8%). In older groups (>50), neoadjuvant radiation was associated with survival benefits in stage II (HR: 0.741; 95% CI: 0.599-0.916; P = 5.80e-3) and stage III (HR: 0.656; 95% CI 0.564-0.764; P = 5.26e-08). Interestingly, neoadjuvant radiation did not increase survival rate in younger patients (< = 50) both in stage II (HR: 2.014; 95% CI: 0.9032-4.490; P = 0.087) and stage III (HR: 1.168; 95% CI: 0.829-1.646; P = 0.372). Additionally, neoadjuvant radiation significantly decreased the cancer-specific mortality in older patients, but increased mortality in younger patients. Our results provided new insights on the neoadjuvant radiation in rectal cancer, especially for the younger patients.

Global Prioritizing Disease Candidate lncRNAs via a Multi-level Composite Network.

LncRNAs play pivotal roles in many important biological processes, but research on the functions of lncRNAs in human disease is still in its infancy. Therefore, it is urgent to prioritize lncRNAs that are potentially associated with diseases. In this work, we developed a novel algorithm, LncPriCNet, that uses a multi-level composite network to prioritize candidate lncRNAs associated with diseases. By integrating genes, lncRNAs, phenotypes and their associations, LncPriCNet achieves an overall performance superior to that of previous methods, with high AUC values of up to 0.93. Notably, LncPriCNet still performs well when information on known disease lncRNAs is lacking. When applied to breast cancer, LncPriCNet identified known breast cancer-related lncRNAs, revealed novel lncRNA candidates and inferred their functions via pathway analysis. We further constructed the human disease-lncRNA landscape, revealed the modularity of the disease-lncRNA network and identified several lncRNA hotspots. In summary, LncPriCNet is a useful tool for prioritizing disease-related lncRNAs and may facilitate understanding of the molecular mechanisms of human disease at the lncRNA level.