CGMega: explainable graph neural network framework with attention mechanisms for cancer gene module dissection.

Cancer is rarely the straightforward consequence of an abnormality in a single gene, but rather reflects a complex interplay of many genes, represented as gene modules. Here, we leverage the recent advances of model-agnostic interpretation approach and develop CGMega, an explainable and graph attention-based deep learning framework to perform cancer gene module dissection. CGMega outperforms current approaches in cancer gene prediction, and it provides a promising approach to integrate multi-omics information. We apply CGMega to breast cancer cell line and acute myeloid leukemia (AML) patients, and we uncover the high-order gene module formed by ErbB family and tumor factors NRG1, PPM1A and DLG2. We identify 396 candidate AML genes, and observe the enrichment of either known AML genes or candidate AML genes in a single gene module. We also identify patient-specific AML genes and associated gene modules. Together, these results indicate that CGMega can be used to dissect cancer gene modules, and provide high-order mechanistic insights into cancer development and heterogeneity.

CLIC3 interacts with NAT10 to inhibit N4-acetylcytidine modification of p21 mRNA and promote bladder cancer progression.

Chromatin accessibility plays important roles in revealing the regulatory networks of gene expression, while its application in bladder cancer is yet to be fully elucidated. Chloride intracellular channel 3 (CLIC3) protein has been reported to be associated with the progression of some tumors, whereas the specific mechanism of CLIC3 in tumor remains unclear. Here, we screened for key genes in bladder cancer through the identification of transcription factor binding site clustered region (TFCR) on the basis of chromatin accessibility and TF motif. CLIC3 was identified by joint profiling of chromatin accessibility data with TCGA database. Clinically, CLIC3 expression was significantly elevated in bladder cancer and was negatively correlated with patient survival. CLIC3 promoted the proliferation of bladder cancer cells by reducing p21 expression in vitro and in vivo. Mechanistically, CLIC3 interacted with NAT10 and inhibited the function of NAT10, resulting in the downregulation of ac4C modification and stability of p21 mRNA. Overall, these findings uncover an novel mechanism of mRNA ac4C modification and CLIC3 may act as a potential therapeutic target for bladder cancer.

Comparative study on genomic and epigenomic profiles of retinoblastoma or tuberous sclerosis complex via nanopore sequencing and a joint screening framework.

Recurrence and extraocular metastasis in advanced intraocular retinoblastoma (RB) are still major obstacles for successful treatment of Chinese children. Tuberous sclerosis complex (TSC) is a very rare, multisystemic genetic disorder characterized by hamartomatous growth. In this study, we aimed to compare genomic and epigenomic profiles with human RB or TSC using recently developed nanopore sequencing, and to identify disease-associated variations or genes. Peripheral blood samples were collected from either RB or RB/TSC patients plus their normal siblings, followed by nanopore sequencing and identification of disease-specific structural variations (SVs) and differentially methylated regions (DMRs) by a systematic biology strategy named as multiomics-based joint screening framework. In total, 316 RB- and 1295 TSC-unique SVs were identified, as well as 1072 RB- and 1114 TSC-associated DMRs, respectively. We eventually identified 6 key genes for RB for further functional validation. Knockdown of CDK19 with specific siRNAs significantly inhibited Y79 cellular proliferation and increased sensitivity to carboplatin, whereas downregulation of AHNAK2 promoted the cell growth as well as drug resistance. Those two genes might serve as potential diagnostic markers or therapeutic targets of RB. The systematic biology strategy combined with functional validation might be an effective approach for rare pediatric malignances with limited samples and challenging collection process.

An interpretable artificial intelligence framework for designing synthetic lethality-based anti-cancer combination therapies.

INTRODUCTION: Synthetic lethality (SL) provides an opportunity to leverage different genetic interactions when designing synergistic combination therapies. To further explore SL-based combination therapies for cancer treatment, it is important to identify and mechanistically characterize more SL interactions. Artificial intelligence (AI) methods have recently been proposed for SL prediction, but the results of these models are often not interpretable such that deriving the underlying mechanism can be challenging. OBJECTIVES: This study aims to develop an interpretable AI framework for SL prediction and subsequently utilize it to design SL-based synergistic combination therapies. METHODS: We propose a knowledge and data dual-driven AI framework for SL prediction (KDDSL). Specifically, we use gene knowledge related to the SL mechanism to guide the construction of the model and develop a method to identify the most relevant gene knowledge for the predicted results. RESULTS: Experimental and literature-based validation confirmed a good balance between predictive and interpretable ability when using KDDSL. Moreover, we demonstrated that KDDSL could help to discover promising drug combinations and clarify associated biological processes, such as the combination of MDM2 and CDK9 inhibitors, which exhibited significant anti-cancer effects in vitro and in vivo. CONCLUSION: These data underscore the potential of KDDSL to guide SL-based combination therapy design. There is a need for biomedicine-focused AI strategies to combine rational biological knowledge with developed models.

COMMO: a web server for the identification and analysis of consensus gene modules across multiple methods.

SUMMARY: A variety of computational methods have been developed to identify functionally related gene modules from genome-wide gene expression profiles. Integrating the results of these methods to identify consensus modules is a promising approach to produce more accurate and robust results. In this application note, we introduce COMMO, the first web server to identify and analyze consensus gene functionally related gene modules from different module detection methods. First, COMMO implements eight state-of-the-art module detection methods and two consensus clustering algorithms. Second, COMMO provides users with mRNA and protein expression data for 33 cancer types from three public databases. Users can also upload their own data for module detection. Third, users can perform functional enrichment and two types of survival analyses on the observed gene modules. Finally, COMMO provides interactive, customizable visualizations and exportable results. With its extensive analysis and interactive capabilities, COMMO offers a user-friendly solution for conducting module-based precision medicine research. AVAILABILITY AND IMPLEMENTATION: COMMO web is available at https://commo.ncpsb.org.cn/, with the source code available on GitHub: https://github.com/Song-xinyu/COMMO/tree/master.

PC3T: a signature-driven predictor of chemical compounds for cellular transition.

Cellular transitions hold great promise in translational medicine research. However, therapeutic applications are limited by the low efficiency and safety concerns of using transcription factors. Small molecules provide a temporal and highly tunable approach to overcome these issues. Here, we present PC3T, a computational framework to enrich molecules that induce desired cellular transitions, and PC3T was able to consistently enrich small molecules that had been experimentally validated in both bulk and single-cell datasets. We then predicted small molecule reprogramming of fibroblasts into hepatic progenitor-like cells (HPLCs). The converted cells exhibited epithelial cell-like morphology and HPLC-like gene expression pattern. Hepatic functions were also observed, such as glycogen storage and lipid accumulation. Finally, we collected and manually curated a cell state transition resource containing 224 time-course gene expression datasets and 153 cell types. Our framework, together with the data resource, is freely available at http://pc3t.idrug.net.cn/ . We believe that PC3T is a powerful tool to promote chemical-induced cell state transitions.

MOCSS: Multi-omics data clustering and cancer subtyping via shared and specific representation learning.

Cancer is an extremely complex disease and each type of cancer usually has several different subtypes. Multi-omics data can provide more comprehensive biological information for identifying and discovering cancer subtypes. However, existing unsupervised cancer subtyping methods cannot effectively learn comprehensive shared and specific information of multi-omics data. Therefore, a novel method is proposed based on shared and specific representation learning. For each omics data, two autoencoders are applied to extract shared and specific information, respectively. To reduce redundancy and mutual interference, orthogonality constraint is introduced to separate shared and specific information. In addition, contrastive learning is applied to align the shared information and strengthen their consistency. Finally, the obtained shared and specific information for all samples are used for clustering tasks to achieve cancer subtyping. Experimental results demonstrate that the proposed method can effectively capture shared and specific information of multi-omics data and outperform other state-of-the-art methods on cancer subtyping.

A Genetic Algorithm-Based Ensemble Learning Framework for Drug Combination Prediction.

Combination therapy is a promising clinical treatment strategy for cancer and other complex diseases. Multiple drugs can target multiple proteins and pathways, greatly improving the therapeutic effect and slowing down drug resistance. To narrow the search space of synergistic drug combinations, many prediction models have been developed. However, drug combination datasets always have the characteristics of class imbalance. Synergistic drug combinations receive the most attention in clinical application but are in small numbers. To predict synergistic drug combinations in different cancer cell lines, in this study, we propose a genetic algorithm-based ensemble learning framework, GA-DRUG, to address the problems of class imbalance and high dimensionality of input data. The cell-line-specific gene expression profiles under drug perturbations are used to train GA-DRUG, which contains imbalanced data processing and the search of global optimal solutions. Compared to 11 state-of-the-art algorithms, GA-DRUG achieves the best performance and significantly improves the prediction performance in the minority class (Synergy). The ensemble framework can effectively correct the classification results of a single classifier. In addition, the cellular proliferation experiment performed on several previously unexplored drug combinations further confirms the predictive ability of GA-DRUG.

A graph neural network-based interpretable framework reveals a novel DNA fragility-associated chromatin structural unit.

BACKGROUND: DNA double-strand breaks (DSBs) are among the most deleterious DNA lesions, and they can cause cancer if improperly repaired. Recent chromosome conformation capture techniques, such as Hi-C, have enabled the identification of relationships between the 3D chromatin structure and DSBs, but little is known about how to explain these relationships, especially from global contact maps, or their contributions to DSB formation. RESULTS: Here, we propose a framework that integrates graph neural network (GNN) to unravel the relationship between 3D chromatin structure and DSBs using an advanced interpretable technique GNNExplainer. We identify a new chromatin structural unit named the DNA fragility-associated chromatin interaction network (FaCIN). FaCIN is a bottleneck-like structure, and it helps to reveal a universal form of how the fragility of a piece of DNA might be affected by the whole genome through chromatin interactions. Moreover, we demonstrate that neck interactions in FaCIN can serve as chromatin structural determinants of DSB formation. CONCLUSIONS: Our study provides a more systematic and refined view enabling a better understanding of the mechanisms of DSB formation under the context of the 3D genome.

A hybrid deep forest-based method for predicting synergistic drug combinations.

Combination therapy is a promising approach in treating multiple complex diseases. However, the large search space of available drug combinations exacerbates challenge for experimental screening. To predict synergistic drug combinations in different cancer cell lines, we propose an improved deep forest-based method, ForSyn, and design two forest types embedded in ForSyn. ForSyn handles imbalanced and high-dimensional data in medium-/small-scale datasets, which are inherent characteristics of drug combination datasets. Compared with 12 state-of-the-art methods, ForSyn ranks first on four metrics for eight datasets with different feature combinations. We conduct a systematic analysis to identify the most appropriate configuration parameters. We validate the predictive value of ForSyn with cell-based experiments on several previously unexplored drug combinations. Finally, a systematic analysis of feature importance is performed on the top contributing features extracted by ForSyn. The resulting key genes may play key roles on corresponding cancers.

Decitabine priming increases anti-PD-1 antitumor efficacy by promoting CD8+ progenitor exhausted T cell expansion in tumor models.

CD8+ exhausted T cells (Tex) are heterogeneous. PD-1 inhibitors reinvigorate progenitor Tex, which subsequently differentiate into irresponsive terminal Tex. The ability to maintain a capacity for durable proliferation of progenitor Tex is important, but the mechanism remains unclear. Here, we showed CD8+ progenitor Tex pretreated with decitabine, a low-dose DNA demethylating agent, had enhanced proliferation and effector function against tumors after anti-PD-1 treatment in vitro. Treatment with decitabine plus anti-PD-1 promoted the activation and expansion of tumor-infiltrated CD8+ progenitor Tex and efficiently suppressed tumor growth in multiple tumor models. Transcriptional and epigenetic profiling of tumor-infiltrated T cells demonstrated that the combination of decitabine plus anti-PD-1 markedly elevated the clonal expansion and cytolytic activity of progenitor Tex compared with anti-PD-1 monotherapy and restrained CD8+ T cell terminal differentiation. Strikingly, decitabine plus anti-PD-1 sustained the expression and activity of the AP-1 transcription factor JunD, which was reduced following PD-1 blockade therapy. Downregulation of JunD repressed T cell proliferation, and activation of JNK/AP-1 signaling in CD8+ T cells enhanced the antitumor capacity of PD-1 inhibitors. Together, epigenetic agents remodel CD8+ progenitor Tex populations and improve responsiveness to anti-PD-1 therapy.

DEML: Drug Synergy and Interaction Prediction Using Ensemble-Based Multi-Task Learning.

Synergistic drug combinations have demonstrated effective therapeutic effects in cancer treatment. Deep learning methods accelerate identification of novel drug combinations by reducing the search space. However, potential adverse drug-drug interactions (DDIs), which may increase the risks for combination therapy, cannot be detected by existing computational synergy prediction methods. We propose DEML, an ensemble-based multi-task neural network, for the simultaneous optimization of five synergy regression prediction tasks, synergy classification, and DDI classification tasks. DEML uses chemical and transcriptomics information as inputs. DEML adapts the novel hybrid ensemble layer structure to construct higher order representation using different perspectives. The task-specific fusion layer of DEML joins representations for each task using a gating mechanism. For the Loewe synergy prediction task, DEML overperforms the state-of-the-art synergy prediction method with an improvement of 7.8% and 13.2% for the root mean squared error and the R2 correlation coefficient. Owing to soft parameter sharing and ensemble learning, DEML alleviates the multi-task learning 'seesaw effect' problem and shows no performance loss on other tasks. DEML has a superior ability to predict drug pairs with high confidence and less adverse DDIs. DEML provides a promising way to guideline novel combination therapy strategies for cancer treatment.

GADRP: graph convolutional networks and autoencoders for cancer drug response prediction.

Drug response prediction in cancer cell lines is of great significance in personalized medicine. In this study, we propose GADRP, a cancer drug response prediction model based on graph convolutional networks (GCNs) and autoencoders (AEs). We first use a stacked deep AE to extract low-dimensional representations from cell line features, and then construct a sparse drug cell line pair (DCP) network incorporating drug, cell line, and DCP similarity information. Later, initial residual and layer attention-based GCN (ILGCN) that can alleviate over-smoothing problem is utilized to learn DCP features. And finally, fully connected network is employed to make prediction. Benchmarking results demonstrate that GADRP can significantly improve prediction performance on all metrics compared with baselines on five datasets. Particularly, experiments of predictions of unknown DCP responses, drug-cancer tissue associations， and drug-pathway associations illustrate the predictive power of GADRP. All results highlight the effectiveness of GADRP in predicting drug responses, and its potential value in guiding anti-cancer drug selection.

Ionizing radiation damage and repair from 3D-genomic perspective.

Ionizing radiation (IR)-induced DNA damage and repair are complex and occur at hierarchical chromatin structures; radiobiology needs to be studied from a 3D-genomic perspective. Differences in IR damage and repair throughout the 3D genome may help to explain differences in radiosensitivity.

Recurrent RNA edits in human preimplantation potentially enhance maternal mRNA clearance.

Posttranscriptional modification plays an important role in key embryonic processes. Adenosine-to-inosine RNA editing, a common example of such modifications, is widespread in human adult tissues and has various functional impacts and clinical consequences. However, whether it persists in a consistent pattern in most human embryos, and whether it supports embryonic development, are poorly understood. To address this problem, we compiled the largest human embryonic editome from 2,071 transcriptomes and identified thousands of recurrent embryonic edits (>=50% chances of occurring in a given stage) for each early developmental stage. We found that these recurrent edits prefer exons consistently across stages, tend to target genes related to DNA replication, and undergo organized loss in abnormal embryos and embryos from elder mothers. In particular, these recurrent edits are likely to enhance maternal mRNA clearance, a possible mechanism of which could be introducing more microRNA binding sites to the 3'-untranslated regions of clearance targets. This study suggests a potentially important, if not indispensable, role of RNA editing in key human embryonic processes such as maternal mRNA clearance; the identified editome can aid further investigations.

NeRD: a multichannel neural network to predict cellular response of drugs by integrating multidimensional data.

BACKGROUND: Considering the heterogeneity of tumors, it is a key issue in precision medicine to predict the drug response of each individual. The accumulation of various types of drug informatics and multi-omics data facilitates the development of efficient models for drug response prediction. However, the selection of high-quality data sources and the design of suitable methods remain a challenge. METHODS: In this paper, we design NeRD, a multidimensional data integration model based on the PRISM drug response database, to predict the cellular response of drugs. Four feature extractors, including drug structure extractor (DSE), molecular fingerprint extractor (MFE), miRNA expression extractor (mEE), and copy number extractor (CNE), are designed for different types and dimensions of data. A fully connected network is used to fuse all features and make predictions. RESULTS: Experimental results demonstrate the effective integration of the global and local structural features of drugs, as well as the features of cell lines from different omics data. For all metrics tested on the PRISM database, NeRD surpassed previous approaches. We also verified that NeRD has strong reliability in the prediction results of new samples. Moreover, unlike other algorithms, when the amount of training data was reduced, NeRD maintained stable performance. CONCLUSIONS: NeRD's feature fusion provides a new idea for drug response prediction, which is of great significance for precise cancer treatment.

Status and Determinants of Symptoms of Anxiety and Depression among Food Delivery Drivers in Shanghai, China.

(1) Background: The psychological status of employees, especially vulnerable populations, has received considerable research attention. However, as a newly emerging and popular occupation in the gig industry, food delivery drivers have received little attention. The majority of these workers are immigrants who are already in a precarious position due to a lack of available jobs, inadequate medical care, poor diets, and communication and acculturation difficulties even before they take these jobs, which involve long working hours and exposure to the elements. (2) Methods: To examine the anxiety and depression symptoms of these workers and possible influencing factors, a cross-sectional study was conducted with a sample of food delivery drivers working for the Meituan Company (one of the largest e-platform companies in China). Anxiety and depression scales were adapted from the GAD-7, and the PHQ-9 was used to assess participants' related symptoms. Differences were compared in terms of sociodemographic, work situation, and lifestyle variables. Binary logistic regressions were conducted to analyze the effects of various factors on the two psychological dimensions. (3) Results: Among the 657 participants, the proportions of participants reporting anxiety and depression symptoms were 46.0% and 18.4%, respectively. Lack of communication with leaders (ORAN = 2.620, 95% CI: 1.528-4.493, p < 0.001; ORDE = 1.928, 95% CI: 1.039-3.577, p = 0.037) and poor sleep quality (ORAN = 2.152, 95% CI: 1.587-2.917, p < 0.001; ORDE = 2.420, 95% CI: 1.672-3.504, p < 0.001) were significant risk factors for both anxiety and depression symptoms. Women (OR = 2.679, 95% CI: 1.621-4.427, p < 0.001), those who climbed ≥31 floors per day (OR = 2.415, 95% CI: 1.189-4.905, p = 0.015), and those with a high frequency of breakfast consumption (OR = 3.821, 95% CI: 1.284-11.369, p = 0.016) were more likely to have anxiety symptoms. Participants who earned less than 5000 RMB (OR = 0.438, 95% CI: 0.204-0.940, p = 0.034), were unwilling to seek medical help (OR = 3.549, 95% CI: 1.846-6.821, p < 0.001), or had a high frequency of smoking (OR = 5.107, 95% CI: 1.187-21.981, p = 0.029) were more likely to be depressive. (4) Conclusion: The existence of communication channels with leaders and good sleep quality are protective factors for anxiety and depression symptoms. Participants who were female, climbed ≥31floors per day, and had a high frequency of eating breakfast were more likely to have anxiety symptoms, while earning less, unwillingness to seek medical help, and a high frequency of smoking were risk factors for depression symptoms.

A benchmark study of deep learning-based multi-omics data fusion methods for cancer.

BACKGROUND: A fused method using a combination of multi-omics data enables a comprehensive study of complex biological processes and highlights the interrelationship of relevant biomolecules and their functions. Driven by high-throughput sequencing technologies, several promising deep learning methods have been proposed for fusing multi-omics data generated from a large number of samples. RESULTS: In this study, 16 representative deep learning methods are comprehensively evaluated on simulated, single-cell, and cancer multi-omics datasets. For each of the datasets, two tasks are designed: classification and clustering. The classification performance is evaluated by using three benchmarking metrics including accuracy, F1 macro, and F1 weighted. Meanwhile, the clustering performance is evaluated by using four benchmarking metrics including the Jaccard index (JI), C-index, silhouette score, and Davies Bouldin score. For the cancer multi-omics datasets, the methods' strength in capturing the association of multi-omics dimensionality reduction results with survival and clinical annotations is further evaluated. The benchmarking results indicate that moGAT achieves the best classification performance. Meanwhile, efmmdVAE, efVAE, and lfmmdVAE show the most promising performance across all complementary contexts in clustering tasks. CONCLUSIONS: Our benchmarking results not only provide a reference for biomedical researchers to choose appropriate deep learning-based multi-omics data fusion methods, but also suggest the future directions for the development of more effective multi-omics data fusion methods. The deep learning frameworks are available at https://github.com/zhenglinyi/DL-mo .

CAD v1.0: Cancer Antigens Database Platform for Cancer Antigen Algorithm Development and Information Exploration.

Cancer vaccines have gradually attracted attention for their tremendous preclinical and clinical performance. With the development of next-generation sequencing technologies and related algorithms, pipelines based on sequencing and machine learning methods have become mainstream in cancer antigen prediction; of particular focus are neoantigens, mutation peptides that only exist in tumor cells that lack central tolerance and have fewer side effects. The rapid prediction and filtering of neoantigen peptides are crucial to the development of neoantigen-based cancer vaccines. However, due to the lack of verified neoantigen datasets and insufficient research on the properties of neoantigens, neoantigen prediction algorithms still need to be improved. Here, we recruited verified cancer antigen peptides and collected as much relevant peptide information as possible. Then, we discussed the role of each dataset for algorithm improvement in cancer antigen research, especially neoantigen prediction. A platform, Cancer Antigens Database (CAD, http://cad.bio-it.cn/), was designed to facilitate users to perform a complete exploration of cancer antigens online.

ExpressVis: a biologist-oriented interactive web server for exploring multi-omics data.

In the era of life-omics, huge amounts of multi-omics data have been generated and widely used in biomedical research. It is challenging for biologists with limited programming skills to obtain biological insights from multi-omics data. Thus, a biologist-oriented platform containing visualization functions is needed to make complex omics data digestible. Here, we propose an easy-to-use, interactive web server named ExpressVis. In ExpressVis, users can prepare datasets; perform differential expression analysis, clustering analysis, and survival analysis; and integrate expression data with protein-protein interaction networks and pathway maps. These analyses are organized into six modules. Users can use each module independently or use several modules interactively. ExpressVis displays analysis results in interactive figures and tables, and provides comprehensive interactive operations in each figure and table, between figures or tables in each module, and among different modules. It is freely accessible at https://omicsmining.ncpsb.org.cn/ExpressVis and does not require login. To test the performance of ExpressVis for multi-omics studies of clinical cohorts, we re-analyzed a published hepatocellular carcinoma dataset and reproduced their main findings, suggesting that ExpressVis is convenient enough to analyze multi-omics data. Based on its complete analysis processes and unique interactive operations, ExpressVis provides an easy-to-use solution for exploring multi-omics data.