The Integrative Method Based on the Module-Network for Identifying Driver Genes in Cancer Subtypes

MPASL: multi-perspective learning knowledge graph attention network for synthetic lethality prediction in human cancer

Synthetic lethality (SL) is widely used to discover the anti-cancer drug targets. However, the identification of SL interactions through wet experiments is costly and inefficient. Hence, the development of efficient and high-accuracy computational methods for SL interactions prediction is of great significance. In this study, we propose MPASL, a multi-perspective learning knowledge graph attention network to enhance synthetic lethality prediction. MPASL utilizes knowledge graph hierarchy propagation to explore multi-source neighbor nodes related to genes. The knowledge graph ripple propagation expands gene representations through existing gene SL preference sets. MPASL can learn the gene representations from both gene-entity perspective and entity-entity perspective. Specifically, based on the aggregation method, we learn to obtain gene-oriented entity embeddings. Then, the gene representations are refined by comparing the various layer-wise neighborhood features of entities using the discrepancy contrastive technique. Finally, the learned gene representation is applied in SL prediction. Experimental results demonstrated that MPASL outperforms several state-of-the-art methods. Additionally, case studies have validated the effectiveness of MPASL in identifying SL interactions between genes.

Privacy-Preserving Identification of Cancer Subtype-Specific Driver Genes Based on Multigenomics Data with Privatedriver

Identifying cancer subtype-specific driver genes from a large number of irrelevant passengers is crucial for targeted therapy in cancer treatment. Recently, the rapid accumulation of large-scale cancer genomics data from multiple institutions has presented remarkable opportunities for identification of cancer subtype-specific driver genes. However, the insufficient subtype samples, privacy issues, and heterogenous of aberration events pose great challenges in precisely identifying cancer subtype-specific driver genes. To address this, we introduce privatedriver, the first model for identifying subtype-specific driver genes that integrates genomics data from multiple institutions in a data privacy-preserving collaboration manner. The process of identifying subtype-specific cancer driver genes using privatedriver involves the following two steps: genomics data integration and collaborative training. In the integration process, the aberration events from multiple genomics data sources are combined for each institution using the forward and backward propagation method of NetICS. In the collaborative training process, each institution utilizes the federated learning framework to upload encrypted model parameters instead of raw data of all institutions to train a global model by using the non-negative matrix factorization algorithm. We applied privatedriver on head and neck squamous cell and colon cancer from The Cancer Genome Atlas website and evaluated it with two benchmarks using macro-Fscore. The comparison analysis demonstrates that privatedriver achieves comparable results to centralized learning models and outperforms most other nonprivacy preserving models, all while ensuring the confidentiality of patient information. We also demonstrate that, for varying predicted driver gene distributions in subtype, our model fully considers the heterogeneity of subtype and identifies subtype-specific driver genes corresponding to the given prognosis and therapeutic effect. The success of privatedriver reveals the feasibility and effectiveness of identifying cancer subtype-specific driver genes in a data protection manner, providing new insights for future privacy-preserving driver gene identification studies.

Advances in computational methods for identifying cancer driver genes

Cancer driver genes (CDGs) are crucial in cancer prevention, diagnosis and treatment. This study employed computational methods for identifying CDGs, categorizing them into four groups. The major frameworks for each of these four categories were summarized. Additionally, we systematically gathered data from public databases and biological networks, and we elaborated on computational methods for identifying CDGs using the aforementioned databases. Further, we summarized the algorithms, mainly involving statistics and machine learning, used for identifying CDGs. Notably, the performances of nine typical identification methods for eight types of cancer were compared to analyze the applicability areas of these methods. Finally, we discussed the challenges and prospects associated with methods for identifying CDGs. The present study revealed that the network-based algorithms and machine learning-based methods demonstrated superior performance.

MAGCN: A Multiple Attention Graph Convolution Networks for Predicting Synthetic Lethality

Synthetic lethality (SL) is a potential cancer therapeutic strategy and drug discovery. Computational approaches to identify synthetic lethality genes have become an effective complement to wet experiments which are time consuming and costly. Graph convolutional networks (GCN) has been utilized to such prediction task as be good at capturing the neighborhood dependency in a graph. However, it is still a lack of the mechanism of aggregating the complementary neighboring information from various heterogeneous graphs. Here, we propose the Multiple Attention Graph Convolution Networks for predicting synthetic lethality (MAGCN). First, we obtain the functional similarity features and topological structure features of genes from different data sources respectively, such as Gene Ontology data and Protein-Protein Interaction. Then, graph convolutional network is utilized to accumulate the knowledge from neighbor nodes according to synthetic lethal associations. Meanwhile, we propose a multiple graphs attention model and construct a multiple graphs attention network to learn the contribution factors of different graphs to generate embedded representation by aggregating these graphs. Finally, the generated feature matrix is decoded to predict potential synthetic lethal interaction. Experimental results show that MAGCN is superior to other baseline methods. Case study demonstrates the ability of MAGCN to predict human SL gene pairs.

Predicting miRNA-Disease Associations via Combining Probability Matrix Feature Decomposition With Neighbor Learning

Predicting the associations of miRNAs and diseases may uncover the causation of various diseases. Many methods are emerging to tackle the sparse and unbalanced disease related miRNA prediction. Here, we propose a Probabilistic matrix decomposition combined with neighbor learning to identify MiRNA-Disease Associations utilizing heterogeneous data(PMDA). First, we build similarity networks for diseases and miRNAs, respectively, by integrating semantic information and functional interactions. Second, we construct a neighbor learning model in which the neighbor information of individual miRNA or disease is utilized to enhance the association relationship to tackle the spare problem. Third, we predict the potential association between miRNAs and diseases via probability matrix decomposition. The experimental results show that PMDA is superior to other five methods in sparse and unbalanced data. The case study shows that the new miRNA-disease interactions predicted by the PMDA are effective and the performance of the PMDA is superior to other methods.

Multiclass Cancer Prediction Based on Copy Number Variation Using Deep Learning

DNA copy number variation (CNV) is the type of DNA variation which is associated with various human diseases. CNV ranges in size from 1 kilobase to several megabases on a chromosome. Most of the computational research for cancer classification is traditional machine learning based, which relies on handcrafted extraction and selection of features. To the best of our knowledge, the deep learning-based research also uses the step of feature extraction and selection. To understand the difference between multiple human cancers, we developed three end-to-end deep learning models, i.e., DNN (fully connected), CNN (convolution neural network), and RNN (recurrent neural network), to classify six cancer types using the CNV data of 24,174 genes. The strength of an end-to-end deep learning model lies in representation learning (automatic feature extraction). The purpose of proposing more than one model is to find which architecture among them performs better for CNV data. Our best model achieved 92% accuracy with an ROC of 0.99, and we compared the performances of our proposed models with state-of-the-art techniques. Our models have outperformed the state-of-the-art techniques in terms of accuracy, precision, and ROC. In the future, we aim to work on other types of cancers as well.

frDriver: A Functional Region Driver Identification for Protein Sequence

Identifying cancer drivers is a crucial challenge to explain the underlying mechanisms of cancer development. There are many methods to identify cancer drivers based on the single mutation site or the entire gene. But they ignore a large number of functional elements with medium in size. It is hypothesized that mutations occurring in different regions of the protein sequence have different effects on the progression of cancer. Here, we develop a novel functional region driver(frDriver) identification method based on Bayesian probability and multiple linear regression models to identify protein regions that can regulate gene expression levels and have high functional impact potential. Combining gene expression data and somatic mutation data, with functional impact scores(SIFT, PROVEAN) as a priori knowledge, we identified cancer driver regions that are most accurate in predicting gene expression levels. We evaluated the performance of frDriver on the BRCA and GBM datasets from TCGA. The results showed that frDriver identified known cancer drivers and outperformed the other three state-of-the-art methods(eDriver, ActiveDriver and OncodriveCLUST). In addition, we performed KEGG pathway and GO term enrichment analysis, and the results indicated that the cancer drivers predicted by frDriver were related to processes such as cancer formation and gene regulation.

Network analysis with biological data of cancer patients: A scoping review

BACKGROUND & OBJECTIVE

Network Analysis (NA) is a mathematical method that allows exploring relations between units and representing them as a graph. Although NA was initially related to social sciences, the past two decades was introduced in Bioinformatics. The recent growth of the networks' use in biological data analysis reveals the need to further investigate this area. In this work, we attempt to identify the use of NA with biological data, and specifically: (a) what types of data are used and whether they are integrated or not, (b) what is the purpose of this analysis, predictive or descriptive, and (c) the outcome of such analyses, specifically in cancer diseases.

METHODS & MATERIALS

The literature review was conducted on two databases, PubMed & IEEE, and was restricted to journal articles of the last decade (January 2010 - December 2019). At a first level, all articles were screened by title and abstract, and at a second level the screening was conducted by reading the full text article, following the predefined inclusion & exclusion criteria leading to 131 articles of interest. A table was created with the information of interest and was used for the classification of the articles. The articles were initially classified to analysis studies and studies that propose a new algorithm or methodology. Each one of these categories was further screened by the following clustering criteria: (a) data used, (b) study purpose, (c) study outcome. Specifically for the studies proposing a new algorithm, the novelty presented in each one was detected.

RESULTS

& Conclusions: In the past five years researchers are focusing on creating new algorithms and methodologies to enhance this field. The articles' classification revealed that only 25% of the analyses are integrating multi-omics data, although 50% of the new algorithms developed follow this integrative direction. Moreover, only 20% of the analyses and 10% of the newly developed methodologies have a predictive purpose. Regarding the result of the works reviewed, 75% of the studies focus on identifying, prognostic or not, gene signatures. Concluding, this review revealed the need for deploying predictive and multi-omics integrative algorithms and methodologies that can be used to enhance cancer diagnosis, prognosis and treatment.

A novel method to identify gene interaction patterns

BACKGROUND

Gene interaction patterns, including modules and motifs, can be used to identify cancer specific biomarkers and to reveal the mechanism of tumorigenesis. Most of the existing module network inferencing methods focus on gene independent functional patterns, while the studies of overlapping characteristics between modules are lacking. The objective of this study was to reveal the functional overlapping patterns in gene modules, helping elucidate the regulatory relationship between overlapping genes and communities, as well as to explore cancer formation and progression.

RESULTS

We analyzed six cancer datasets from The Cancer Genome Atlas and obtained three kinds of gene functional modules for each cancer, including Independent-Community, Dependent-Community and Merged-Community. In the six cancers, 59(3.5%) Independent-Communities were identified, while 1631(96.5%) Dependent-Communities were acquired. Compared with Lemon-Tree and K-Means, the gene communities identified by our method were enriched in more known GO categories with lower p-values. Meanwhile, those identified distinguishing communities can significantly distinguish the survival prognostic of patients by Kaplan-Meier analysis. Furthermore, identified driver genes in the gene communities can be considered as biomarkers which can accurately distinguish the tumour or normal samples for each cancer type.

CONCLUSIONS

In all identified communities, Dependent-Communities are the majority. Our method is more effective than the other two methods which do not consider the overlapping characteristics of modules. This indicates that overlapping genes are located in different specific functional groups, and a communication bridge is established between the communities to construct a comprehensive carcinogenesis.

Evaluating machine learning methodologies for identification of cancer driver genes

Cancer is driven by distinctive sorts of changes and basic variations in genes. Recognizing cancer driver genes is basic for accurate oncological analysis. Numerous methodologies to distinguish and identify drivers presently exist, but efficient tools to combine and optimize them on huge datasets are few. Most strategies for prioritizing transformations depend basically on frequency-based criteria. Strategies are required to dependably prioritize organically dynamic driver changes over inert passengers in high-throughput sequencing cancer information sets. This study proposes a model namely PCDG-Pred which works as a utility capable of distinguishing cancer driver and passenger attributes of genes based on sequencing data. Keeping in view the significance of the cancer driver genes an efficient method is proposed to identify the cancer driver genes. Further, various validation techniques are applied at different levels to establish the effectiveness of the model and to obtain metrics like accuracy, Mathew's correlation coefficient, sensitivity, and specificity. The results of the study strongly indicate that the proposed strategy provides a fundamental functional advantage over other existing strategies for cancer driver genes identification. Subsequently, careful experiments exhibit that the accuracy metrics obtained for self-consistency, independent set, and cross-validation tests are 91.08%., 87.26%, and 92.48% respectively.

TPQCI: A topology potential-based method to quantify functional influence of copy number variations

Copy number variation (CNV) is a major type of chromosomal structural variation that play important roles in many diseases including cancers. Due to genome instability, a large number of CNV events can be detected in diseases such as cancer. Therefore, it is important to identify the functionally important CNVs in diseases, which currently still poses a challenge in genomics. One of the critical steps to solve the problem is to define the influence of CNV. In this paper, we provide a topology potential based method, TPQCI, to quantify this kind of influence by integrating statistics, gene regulatory associations, and biological function information. We used this metric to detect functionally enriched genes on genomic segments with CNV in breast cancer and multiple myeloma and discovered biological functions influenced by CNV. Our results demonstrate that, by using our proposed TPQCI metric, we can detect disease-specific genes that are influenced by CNVs. Source codes of TPQCI are provided in Github (https://github.com/usos/TPQCI).

sgRNA-PSM: Predict sgRNAs On-Target Activity Based on Position-Specific Mismatch

As a key technique for the CRISPR-Cas9 system, identification of single-guide RNAs (sgRNAs) on-target activity is critical for both theoretical research (investigation of RNA functions) and real-world applications (genome editing and synthetic biology). Because of its importance, several computational predictors have been proposed to predict sgRNAs on-target activity. All of these methods have clearly contributed to the developments of this very important field. However, they are suffering from certain limitations. We proposed two new methods called "sgRNA-PSM" and "sgRNA-ExPSM" for sgRNAs on-target activity prediction via capturing the long-range sequence information and evolutionary information using a new way to reduce the dimension of the feature vector to avoid the risk of overfitting. Rigorous leave-one-gene-out cross-validation on a benchmark dataset with 11 human genes and 6 mouse genes, as well as an independent dataset, indicated that the two new methods outperformed other competing methods. To make it easier for users to use the proposed sgRNA-PSM predictor, we have established a corresponding web server, which is available at http://bliulab.net/sgRNA-PSM/.

A Method for Prediction of Thermophilic Protein Based on Reduced Amino Acids and Mixed Features

The thermostability of proteins is a key factor considered during enzyme engineering, and finding a method that can identify thermophilic and non-thermophilic proteins will be helpful for enzyme design. In this study, we established a novel method combining mixed features and machine learning to achieve this recognition task. In this method, an amino acid reduction scheme was adopted to recode the amino acid sequence. Then, the physicochemical characteristics, auto-cross covariance (ACC), and reduced dipeptides were calculated and integrated to form a mixed feature set, which was processed using correlation analysis, feature selection, and principal component analysis (PCA) to remove redundant information. Finally, four machine learning methods and a dataset containing 500 random observations out of 915 thermophilic proteins and 500 random samples out of 793 non-thermophilic proteins were used to train and predict the data. The experimental results showed that 98.2% of thermophilic and non-thermophilic proteins were correctly identified using 10-fold cross-validation. Moreover, our analysis of the final reserved features and removed features yielded information about the crucial, unimportant and insensitive elements, it also provided essential information for enzyme design.

Developing a Multi-Layer Deep Learning Based Predictive Model to Identify DNA N4-Methylcytosine Modifications

DNA N4-methylcytosine modification (4mC) plays an essential role in a variety of biological processes. Therefore, accurate identification the 4mC distribution in genome-scale is important for systematically understanding its biological functions. In this study, we present Deep4mcPred, a multi-layer deep learning based predictive model to identify DNA N4-methylcytosine modifications. In this predictor, we for the first time integrate residual network and recurrent neural network to build a multi-layer deep learning predictive system. As compared to existing predictors using traditional machine learning, our proposed method has two advantages. First, our deep learning framework does not need to specify the features when training the predictive model. It can automatically learn the high-level features and capture the characteristic specificity of 4mC sites, benefiting to distinguish true 4mC sites from non-4mC sites. On the other hand, our deep learning method outperforms the traditional machine learning predictors in performance by benchmarking comparison, demonstrating that the proposed Deep4mcPred is more effective in the DNA 4mC site prediction. Moreover, via experimental comparison, we found that attention mechanism introduced into the deep learning framework is useful to capture the critical features. Additionally, we develop a webserver implementing the proposed method for the academic use of research community, which is now available at http://server.malab.cn/Deep4mcPred.

DMCM: a Data-adaptive Mutation Clustering Method to identify cancer-related mutation clusters

Motivation

Functional somatic mutations within coding amino acid sequences confer growth advantage in pathogenic process. Most existing methods for identifying cancer-related mutations focus on the single amino acid or the entire gene level. However, gain-of-function mutations often cluster in specific protein regions instead of existing independently in the amino acid sequences. Some approaches for identifying mutation clusters with mutation density on amino acid chain have been proposed recently. But their performance in identification of mutation clusters remains to be improved.

Results

Here we present a Data-adaptive Mutation Clustering Method (DMCM), in which kernel density estimate (KDE) with a data-adaptive bandwidth is applied to estimate the mutation density, to find variable clusters with different lengths on amino acid sequences. We apply this approach in the mutation data of 571 genes in over twenty cancer types from The Cancer Genome Atlas (TCGA). We compare the DMCM with M2C, OncodriveCLUST and Pfam Domain and find that DMCM tends to identify more significant clusters. The cross-validation analysis shows DMCM is robust and cluster cancer type enrichment analysis shows that specific cancer types are enriched for specific mutation clusters.

Availability and implementation

DMCM is written in Python and analysis methods of DMCM are written in R. They are all released online, available through https://github.com/XinguoLu/DMCM.

Supplementary information

Supplementary data are available at Bioinformatics online.

A Novel Method for Identifying the Potential Cancer Driver Genes Based on Molecular Data Integration

The identification of the cancer driver genes is essential for personalized therapy. The mutation frequency of most driver genes is in the middle (2-20%) or even lower range, which makes it difficult to find the driver genes with low-frequency mutations. Other forms of genomic aberrations, such as copy number variations (CNVs) and epigenetic changes, may also reflect cancer progression. In this work, a method for identifying the potential cancer driver genes (iPDG) based on molecular data integration is proposed. DNA copy number variation, somatic mutation, and gene expression data of matched cancer samples are integrated. In combination with the method of iKEEG, the "key genes" of cancer are identified, and the change in their expression levels is used for auxiliary evaluation of whether the mutated genes are potential drivers. For a mutated gene, the concept of mutational effect is defined, which takes into account the effects of copy number variation, mutation gene itself, and its neighbor genes. The method mainly includes two steps: the first step is data preprocessing. First, DNA copy number variation and somatic mutation data are integrated. Then, the integrated data are mapped to a given interaction network, and the diffusion kernel is used to form the mutation effect matrix. The second step is to obtain the key genes by using the iKGGE method, and construct the connection matrix by means of the gene expression data of the key genes and mutation impact matrix of the mutated genes. Experiments on TCGA breast cancer and Glioblastoma multiforme datasets demonstrate that iPDG is effective not only to identify the known cancer driver genes but also to discover the rare potential driver genes. When measured by functional enrichment analysis, we find that these genes are clearly associated with these two types of cancers.