Biological knowledge graph-guided investigation of immune therapy response in cancer with graph neural network.

The determination of transcriptome profiles that mediate immune therapy in cancer remains a major clinical and biological challenge. Despite responses induced by immune-check points inhibitors (ICIs) in diverse tumor types and all the big breakthroughs in cancer immunotherapy, most patients with solid tumors do not respond to ICI therapies. It still remains a big challenge to predict the ICI treatment response. Here, we propose a framework with multiple prior knowledge networks guided for immune checkpoints inhibitors prediction-DeepOmix-ICI (or ICInet for short). ICInet can predict the immune therapy response by leveraging geometric deep learning and prior biological knowledge graphs of gene-gene interactions. Here, we demonstrate more than 600 ICI-treated patients with ICI response data and gene expression profile to apply on ICInet. ICInet was used for ICI therapy responses prediciton across different cancer types-melanoma, gastric cancer and bladder cancer, which includes 7 cohorts from different data sources. ICInet is able to robustly generalize into multiple cancer types. Moreover, the performance of ICInet in those cancer types can outperform other ICI biomarkers in the clinic. Our model [area under the curve (AUC = 0.85)] generally outperformed other measures, including tumor mutational burden (AUC = 0.62) and programmed cell death ligand-1 score (AUC = 0.74). Therefore, our study presents a prior-knowledge guided deep learning method to effectively select immunotherapy-response-associated biomarkers, thereby improving the prediction of immunotherapy response for precision oncology.

Multi-modality artificial intelligence in digital pathology.

In common medical procedures, the time-consuming and expensive nature of obtaining test results plagues doctors and patients. Digital pathology research allows using computational technologies to manage data, presenting an opportunity to improve the efficiency of diagnosis and treatment. Artificial intelligence (AI) has a great advantage in the data analytics phase. Extensive research has shown that AI algorithms can produce more up-to-date and standardized conclusions for whole slide images. In conjunction with the development of high-throughput sequencing technologies, algorithms can integrate and analyze data from multiple modalities to explore the correspondence between morphological features and gene expression. This review investigates using the most popular image data, hematoxylin-eosin stained tissue slide images, to find a strategic solution for the imbalance of healthcare resources. The article focuses on the role that the development of deep learning technology has in assisting doctors' work and discusses the opportunities and challenges of AI.

KOBAS-i: intelligent prioritization and exploratory visualization of biological functions for gene enrichment analysis.

Gene set enrichment (GSE) analysis plays an essential role in extracting biological insight from genome-scale experiments. ORA (overrepresentation analysis), FCS (functional class scoring), and PT (pathway topology) approaches are three generations of GSE methods along the timeline of development. Previous versions of KOBAS provided services based on just the ORA method. Here we presented version 3.0 of KOBAS, which is named KOBAS-i (short for KOBAS intelligent version). It introduced a novel machine learning-based method we published earlier, CGPS, which incorporates seven FCS tools and two PT tools into a single ensemble score and intelligently prioritizes the relevant biological pathways. In addition, KOBAS has expanded the downstream exploratory visualization for selecting and understanding the enriched results. The tool constructs a novel view of cirFunMap, which presents different enriched terms and their correlations in a landscape. Finally, based on the previous version's framework, KOBAS increased the number of supported species from 1327 to 5944. For an easier local run, it also provides a prebuilt Docker image that requires no installation, as a supplementary to the source code version. KOBAS can be freely accessed at http://kobas.cbi.pku.edu.cn, and a mirror site is available at http://bioinfo.org/kobas.

HERB: a high-throughput experiment- and reference-guided database of traditional Chinese medicine.

Pharmacotranscriptomics has become a powerful approach for evaluating the therapeutic efficacy of drugs and discovering new drug targets. Recently, studies of traditional Chinese medicine (TCM) have increasingly turned to high-throughput transcriptomic screens for molecular effects of herbs/ingredients. And numerous studies have examined gene targets for herbs/ingredients, and link herbs/ingredients to various modern diseases. However, there is currently no systematic database organizing these data for TCM. Therefore, we built HERB, a high-throughput experiment- and reference-guided database of TCM, with its Chinese name as BenCaoZuJian. We re-analyzed 6164 gene expression profiles from 1037 high-throughput experiments evaluating TCM herbs/ingredients, and generated connections between TCM herbs/ingredients and 2837 modern drugs by mapping the comprehensive pharmacotranscriptomics dataset in HERB to CMap, the largest such dataset for modern drugs. Moreover, we manually curated 1241 gene targets and 494 modern diseases for 473 herbs/ingredients from 1966 references published recently, and cross-referenced this novel information to databases containing such data for drugs. Together with database mining and statistical inference, we linked 12 933 targets and 28 212 diseases to 7263 herbs and 49 258 ingredients and provided six pairwise relationships among them in HERB. In summary, HERB will intensively support the modernization of TCM and guide rational modern drug discovery efforts. And it is accessible through http://herb.ac.cn/.

NONCODEV6: an updated database dedicated to long non-coding RNA annotation in both animals and plants.

NONCODE (http://www.noncode.org/) is a comprehensive database of collection and annotation of noncoding RNAs, especially long non-coding RNAs (lncRNAs) in animals. NONCODEV6 is dedicated to providing the full scope of lncRNAs across plants and animals. The number of lncRNAs in NONCODEV6 has increased from 548 640 to 644 510 since the last update in 2017. The number of human lncRNAs has increased from 172 216 to 173 112. The number of mouse lncRNAs increased from 131 697 to 131 974. The number of plant lncRNAs is 94 697. The relationship between lncRNAs in human and cancer were updated with transcriptome sequencing profiles. Three important new features were also introduced in NONCODEV6: (i) updated human lncRNA-disease relationships, especially cancer; (ii) lncRNA annotations with tissue expression profiles and predicted function in five common plants; iii) lncRNAs conservation annotation at transcript level for 23 plant species. NONCODEV6 is accessible through http://www.noncode.org/.

A robust two-gene signature for glioblastoma survival prediction.

Glioblastoma multiforme (GBM) is a highly malignant brain tumor. We explored the prognostic gene signature in 443 GBM samples by systematic bioinformatics analysis, using GSE16011 with microarray expression and corresponding clinical data from Gene Expression Omnibus as the training set. Meanwhile, patients from The Chinese Glioma Genome Atlas database (CGGA) were used as the test set and The Cancer Genome Atlas database (TCGA) as the validation set. Through Cox regression analysis, Kaplan-Meier analysis, t-distributed Stochastic Neighbor Embedding algorithm, clustering, and receiver operating characteristic analysis, a two-gene signature (GRIA2 and RYR3) associated with survival was selected in the GSE16011 dataset. The GRIA2-RYR3 signature divided patients into two risk groups with significantly different survival in the GSE16011 dataset (median: 0.72, 95% confidence interval [CI]: 0.64-0.98, vs median: 0.98, 95% CI: 0.65-1.61 years, logrank test P < .001), the CGGA dataset (median: 0.84, 95% CI: 0.70-1.18, vs median: 1.21, 95% CI: 0.95-2.94 years, logrank test P = .0017), and the TCGA dataset (median: 1.03, 95% CI: 0.86-1.24, vs median: 1.23, 95% CI: 1.04-1.85 years, logrank test P = .0064), validating the predictive value of the signature. And the survival predictive potency of the signature was independent from clinicopathological prognostic features in multivariable Cox analysis. We found that after transfection of U87 cells with small interfering RNA, GRIA2 and RYR3 influenced the biological behaviors of proliferation, migration, and invasion of glioblastoma cells. In conclusion, the two-gene signature was a robust prognostic model to predict GBM survival.

NONCODEV5: a comprehensive annotation database for long non-coding RNAs.

NONCODE (http://www.bioinfo.org/noncode/) is a systematic database that is dedicated to presenting the most complete collection and annotation of non-coding RNAs (ncRNAs), especially long non-coding RNAs (lncRNAs). Since NONCODE 2016 was released two years ago, the amount of novel identified ncRNAs has been enlarged by the reduced cost of next-generation sequencing, which has produced an explosion of newly identified data. The third-generation sequencing revolution has also offered longer and more accurate annotations. Moreover, accumulating evidence confirmed by biological experiments has provided more comprehensive knowledge of lncRNA functions. The ncRNA data set was expanded by collecting newly identified ncRNAs from literature published over the past two years and integration of the latest versions of RefSeq and Ensembl. Additionally, pig was included in the database for the first time, bringing the total number of species to 17. The number of lncRNAs in NONCODEv5 increased from 527 336 to 548 640. NONCODEv5 also introduced three important new features: (i) human lncRNA-disease relationships and single nucleotide polymorphism-lncRNA-disease relationships were constructed; (ii) human exosome lncRNA expression profiles were displayed; (iii) the RNA secondary structures of NONCODE human transcripts were predicted. NONCODEv5 is also accessible through http://www.noncode.org/.

A four-long noncoding RNA signature predicts survival of hepatocellular carcinoma patients.

BACKGROUND: Hepatocellular carcinoma (HCC) is a common neoplasm located in the liver. Accumulating evidence has highlighted that long noncoding RNAs (lncRNAs) are correlated with the survival of HCC patients. This study focuses on finding a lncRNA signature to predict the prognostic risk of HCC patients. METHODS: Statistical and machine learning analyses were conducted to analyze the lncRNA expression data and corresponding clinical data of 180 HCC patients collected from the public online Tanric and The Cancer Genome Atlas (TCGA) databases. RESULTS: From the training dataset, we obtained the four-lncRNA model comprising RP11-495K9.6, RP11-96O20.2, RP11-359K18.3, and LINC00556 which can divide HCC patients into two different groups with significantly different prognosis (n = 90, median 1.81, 95% confidence interval [CI]: 1.50-4.91 vs 8.56 years, 95% CI: 6.96-9.97, log-rank test P < .001). The test dataset confirmed the prognostic ability of the signature (n = 90, median 1.95, 95% CI: 1.14-4.08 vs 5.80 years, 95% CI: 3.11-6.82, log-rank test P = .007). Receiver operating characteristic curve displayed the better prediction efficiency of the four-lncRNA signature than the tumor/node/metastasis stage. Cox analysis showed the four-lncRNA signature was an independent predictor of HCC prognosis. CONCLUSION: The four-lncRNA signature can be used as an independent biomarker for HCC patients to predict the prognostic risk.

DeepFusion: A deep bimodal information fusion network for unraveling protein-RNA interactions using in vivo RNA structures.

RNA-binding proteins (RBPs) are key post-transcriptional regulators, and the malfunctions of RBP-RNA binding lead to diverse human diseases. However, prediction of RBP binding sites is largely based on RNA sequence features, whereas in vivo RNA structural features based on high-throughput sequencing are rarely incorporated. Here, we designed a deep bimodal information fusion network called DeepFusion for unraveling protein-RNA interactions by incorporating structural features derived from DMS-seq data. DeepFusion integrates two sub-models to extract local motif-like information and long-term context information. We show that DeepFusion performs best compared with other cutting-edge methods with only sequence inputs on two datasets. DeepFusion's performance is further improved with bimodal input after adding in vivo DMS-seq structural features. Furthermore, DeepFusion can be used for analyzing RNA degradation, demonstrating significantly different RBP-binding scores in genes with slow degradation rates versus those with rapid degradation rates. DeepFusion thus provides enhanced abilities for further analysis of functional RNAs. DeepFusion's code and data are available at http://bioinfo.org/deepfusion/.

Predicting transcriptional responses to novel chemical perturbations using deep generative model for drug discovery.

Understanding transcriptional responses to chemical perturbations is central to drug discovery, but exhaustive experimental screening of disease-compound combinations is unfeasible. To overcome this limitation, here we introduce PRnet, a perturbation-conditioned deep generative model that predicts transcriptional responses to novel chemical perturbations that have never experimentally perturbed at bulk and single-cell levels. Evaluations indicate that PRnet outperforms alternative methods in predicting responses across novel compounds, pathways, and cell lines. PRnet enables gene-level response interpretation and in-silico drug screening for diseases based on gene signatures. PRnet further identifies and experimentally validates novel compound candidates against small cell lung cancer and colorectal cancer. Lastly, PRnet generates a large-scale integration atlas of perturbation profiles, covering 88 cell lines, 52 tissues, and various compound libraries. PRnet provides a robust and scalable candidate recommendation workflow and successfully recommends drug candidates for 233 diseases. Overall, PRnet is an effective and valuable tool for gene-based therapeutics screening.

Corrigendum to "TREAT: Therapeutic RNAs exploration inspired by artificial intelligence technology" [Comput Struct Biotechnol J 20 (2022) 5680-5689].

[This corrects the article DOI: 10.1016/j.scr.2023.103115.].

TREAT: Therapeutic RNAs exploration inspired by artificial intelligence technology.

Recent advances in RNA engineering have enabled the development of RNA-based therapeutics for a broad spectrum of applications. Developing RNA therapeutics start with targeted RNA screening and move to the drug design and optimization. However, existing target screening tools ignore noncoding RNAs and their disease-relevant regulatory relationships. And designing therapeutic RNAs encounters high computational complexity of multi-objective optimization to overcome the immunogenicity, instability and inefficient translational production. To unlock the therapeutic potential of noncoding RNAs and enable one-stop screening and design of therapeutic RNAs, we have built the platform TREAT. It incorporates 43,087,953 regulatory relationships between coding and noncoding genes from 81 biological networks under different physiological conditions. TREAT introduces graph representation learning with Random Walk Diffusions to perform disease-relevant target screening, in addition to the commonly utilized Topological Degree and PageRank algorithms. Design and optimization of large RNAs or interfering RNAs are both available. To reduce the computational complexity of multi-objective optimization for large RNA, we stratified the features into local and global features. The local features are evaluated on the fixed-length or dynamic-length local bins, whereas the latter are inspired by AI language models of protein sequence. Then the global assessment is performed on refined candidates, thus reducing the enormous search space. Overall, TREAT is a one-stop platform for the screening and designing of therapeutic RNAs, with particular attention to noncoding RNAs and cutting-edge AI technology embedded, leading the progress of innovative therapeutics for challenging diseases. TREAT is freely accessible at https://rna.org.cn/treat.

DeepOmix: A scalable and interpretable multi-omics deep learning framework and application in cancer survival analysis.

Integrative analysis of multi-omics data can elucidate valuable insights into complex molecular mechanisms for various diseases. However, due to their different modalities and high dimension, utilizing and integrating different types of omics data suffers from great challenges. There is an urgent need to develop a powerful method to improve survival prediction and detect functional gene modules from multi-omics data. To deal with these problems, we present DeepOmix (a scalable and interpretable multi-Omics Deep learning framework and application in cancer survival analysis), a flexible, scalable, and interpretable method for extracting relationships between the clinical survival time and multi-omics data based on a deep learning framework. DeepOmix enables the non-linear combination of variables from different omics datasets and incorporates prior biological information defined by users (such as signaling pathways and tissue networks). Benchmark experiments demonstrate that DeepOmix outperforms the other five cutting-edge prediction methods. Besides, Lower Grade Glioma (LGG) is taken as the case study to perform the prognosis prediction and illustrate the functional module nodes which are associated with the prognostic result in the prediction model.

The long non-coding RNA NONHSAG026900 predicts prognosis as a favorable biomarker in patients with diffuse large B-cell lymphoma.

Long non-coding RNAs are known to be involved in cancer progression, but their biological functions and prognostic values are still largely unexplored in diffuse large B-cell lymphoma. In this study, long non-coding RNAs expression was characterized in 1,403 samples including normal and diffuse large B-cell lymphoma by repurposing 7 microarray datasets. Compared with any stage of normal B cells, NONHSAG026900 expression was significantly decreased in tumor samples. And in germinal center B-cell subtype, the significantly higher expression of NONHSAG026900 indicated it was a favorable prognosis biomarker. Then the prognostic power of NONHSAG026900 was validated with another independent dataset and NONHSAG026900 improved the predictive power of International Prognostic Index as an independent factor. Moreover, functional prediction and validation demonstrated that NONHSAG026900 could inhibit cell cycle activity to restrain tumor proliferation. These findings identified NONHSAG026900 as a novel prognostic biomarker and offered a new therapeutic target for diffuse large B-cell lymphoma patients.