Predicting dynamic signaling network response under unseen perturbations

DNN-PNN: A parallel deep neural network model to improve anticancer drug sensitivity

With the rapid development of deep learning techniques and large-scale genomics database, it is of great potential to apply deep learning to the prediction task of anticancer drug sensitivity, which can effectively improve the identification efficiency and accuracy of therapeutic biomarkers. In this study, we propose a parallel deep learning framework DNN-PNN, which integrates rich and heterogeneous information from gene expression and pharmaceutical chemical structure data. With the proposal of DNN-PNN, a new and more effective drug data representation strategy is introduced, that is, the correlation between features is represented by product, which alleviates the limitations of high-dimensional discrete data in deep learning. Furthermore, the framework is optimized to reduce the time complexity of the model. We conducted extensive experiments on the CCLE datasets to compare DNN-PNN with its variant DNN-FM representing the traditional feature correlation model, the component DNN or PNN alone, and the common machine learning models. It is found that DNN-PNN not only has high prediction accuracy, but also has significant advantages in stability and convergence speed.

Topological estimation of signal flow in complex signaling networks

In a cell, any information about extra- or intra-cellular changes is transferred and processed through a signaling network and dysregulation of signal flow often leads to disease such as cancer. So, understanding of signal flow in the signaling network is critical to identify drug targets. Owing to the development of high-throughput measurement technologies, the structure of a signaling network is becoming more available, but detailed kinetic parameter information about molecular interactions is still very limited. A question then arises as to whether we can estimate the signal flow based only on the structure information of a signaling network. To answer this question, we develop a novel algorithm that can estimate the signal flow using only the topological information and apply it to predict the direction of activity change in various signaling networks. Interestingly, we find that the average accuracy of the estimation algorithm is about 60–80% even though we only use the topological information. We also find that this predictive power gets collapsed if we randomly alter the network topology, showing the importance of network topology. Our study provides a basis for utilizing the topological information of signaling networks in precision medicine or drug target discovery.

A proteogenomic approach to understand splice isoform functions through sequence and expression-based computational modeling

The products of multi-exon genes are a mixture of alternatively spliced isoforms, from which the translated proteins can have similar, different or even opposing functions. It is therefore essential to differentiate and annotate functions for individual isoforms. Computational approaches provide an efficient complement to expensive and time-consuming experimental studies. The input data of these methods range from DNA sequence, to RNA selection pressure, to expressed sequence tags, to full-length complementary DNA, to exon array, to RNA-seq expression, to proteomic data. Notably, RNA-seq technology generates quantitative profiling of transcript expression at the genome scale, with an unprecedented amount of expression data available for developing isoform function prediction methods. Integrative analysis of these data at different molecular levels enables a proteogenomic approach to systematically interrogate isoform functions. Here, we briefly review the state-of-the-art methods according to their input data sources, discuss their advantages and limitations and point out potential ways to improve prediction accuracies.

Predicting Anticancer Drug Responses Using a Dual-Layer Integrated Cell Line-Drug Network Model

The ability to predict the response of a cancer patient to a therapeutic agent is a major goal in modern oncology that should ultimately lead to personalized treatment. Existing approaches to predicting drug sensitivity rely primarily on profiling of cancer cell line panels that have been treated with different drugs and selecting genomic or functional genomic features to regress or classify the drug response. Here, we propose a dual-layer integrated cell line-drug network model, which uses both cell line similarity network (CSN) data and drug similarity network (DSN) data to predict the drug response of a given cell line using a weighted model. Using the Cancer Cell Line Encyclopedia (CCLE) and Cancer Genome Project (CGP) studies as benchmark datasets, our single-layer model with CSN or DSN and only a single parameter achieved a prediction performance comparable to the previously generated elastic net model. When using the dual-layer model integrating both CSN and DSN, our predicted response reached a 0.6 Pearson correlation coefficient with observed responses for most drugs, which is significantly better than the previous results using the elastic net model. We have also applied the dual-layer cell line-drug integrated network model to fill in the missing drug response values in the CGP dataset. Even though the dual-layer integrated cell line-drug network model does not specifically model mutation information, it correctly predicted that BRAF mutant cell lines would be more sensitive than BRAF wild-type cell lines to three MEK1/2 inhibitors tested.

In this study, using the Cancer Cell Line Encyclopedia (CCLE) and Cancer Genome Project (CGP) studies as benchmark datasets, we explored the application of similarity information between cell lines and drugs in drug response prediction. We found that similar cell lines by gene expression profiles exhibit similar response to the same drug. Meanwhile, drugs with similar chemical structures also show similar inhibitory effects across different cell lines. Based on the above observations, we proposed a dual-layer network and local weighted model to predict drug response of a cell line using proximal information of the drug-cell line network. The only three parameters of our model are optimized by leave-one-out cross-validation for each drug. Two case studies of MAPK and ERK signal pathways on CCLE dataset proved that the predicted-to-observed correlations of our dual-layer network model is significantly better than the previous predictor using elastic net model. Interestingly, predictions based on drug similarity network (DSN) alone were much better than those based on cell line similarity network (CSN) alone for most drugs, implying that drug similarities are more informative for drug response prediction than cell line similarities. Our network model can be applied to predict the response of a new cell line to existing already tested drugs or to predict the response of an existing cell line to new drugs, thus potentially saving the cost in a drug-cell line screening.

Functional Networks of Highest-Connected Splice Isoforms: From The Chromosome 17 Human Proteome Project

Alternative splicing allows a single gene to produce multiple transcript-level splice isoforms from which the translated proteins may show differences in their expression and function. Identifying the major functional or canonical isoform is important for understanding gene and protein functions. Identification and characterization of splice isoforms is a stated goal of the HUPO Human Proteome Project and of neXtProt. Multiple efforts have catalogued splice isoforms as "dominant", "principal", or "major" isoforms based on expression or evolutionary traits. In contrast, we recently proposed highest connected isoforms (HCIs) as a new class of canonical isoforms that have the strongest interactions in a functional network and revealed their significantly higher (differential) transcript-level expression compared to nonhighest connected isoforms (NCIs) regardless of tissues/cell lines in the mouse. HCIs and their expression behavior in the human remain unexplored. Here we identified HCIs for 6157 multi-isoform genes using a human isoform network that we constructed by integrating a large compendium of heterogeneous genomic data. We present examples for pairs of transcript isoforms of ABCC3, RBM34, ERBB2, and ANXA7. We found that functional networks of isoforms of the same gene can show large differences. Interestingly, differential expression between HCIs and NCIs was also observed in the human on an independent set of 940 RNA-seq samples across multiple tissues, including heart, kidney, and liver. Using proteomic data from normal human retina and placenta, we showed that HCIs are a promising indicator of expressed protein isoforms exemplified by NUDFB6 and M6PR. Furthermore, we found that a significant percentage (20%, p = 0.0003) of human and mouse HCIs are homologues, suggesting their conservation between species. Our identified HCIs expand the repertoire of canonical isoforms and are expected to facilitate studying main protein products, understanding gene regulation, and possibly evolution. The network is available through our web server as a rich resource for investigating isoform functional relationships (http://guanlab.ccmb.med.umich.edu/hisonet). All MS/MS data were available at ProteomeXchange Web site (http://www.proteomexchange.org) through their identifiers (retina: PXD001242, placenta: PXD000754).

MIsoMine: a genome-scale high-resolution data portal of expression, function and networks at the splice isoform level in the mouse

Products of multiexon genes, especially in higher organisms, are a mixture of isoforms with different or even opposing functions, and therefore need to be treated separately. However, most studies and available resources such as Gene Ontology provide only gene-level function annotations, and therefore lose the differential information at the isoform level. Here we report MIsoMine, a high-resolution portal to multiple levels of functional information of alternatively spliced isoforms in the mouse. This data portal provides tissue-specific expression patterns and co-expression networks, along with such previously published functional genomic data as protein domains, predicted isoform-level functions and functional relationships. The core utility of MIsoMine is allowing users to explore a preprocessed, quality-controlled set of RNA-seq data encompassing diverse tissues and cell lineages. Tissue-specific co-expression networks were established, allowing a 2D ranking of isoforms and tissues by co-expression patterns. The results of the multiple isoforms of the same gene are presented in parallel to facilitate direct comparison, with cross-talking to prioritized functions at the isoform level. MIsoMine provides the first isoform-level resolution effort at genome-scale. We envision that this data portal will be a valuable resource for exploring functional genomic data, and will complement the existing functionalities of the mouse genome informatics database and the gene expression database for the laboratory mouse.

Database URL: http://guanlab.ccmb.med.umich.edu/misomine/

Regulatory network inferred using expression data of small sample size: application and validation in erythroid system

MOTIVATION

Modeling regulatory networks using expression data observed in a differentiation process may help identify context-specific interactions. The outcome of the current algorithms highly depends on the quality and quantity of a single time-course dataset, and the performance may be compromised for datasets with a limited number of samples.

RESULTS

In this work, we report a multi-layer graphical model that is capable of leveraging many publicly available time-course datasets, as well as a cell lineage-specific data with small sample size, to model regulatory networks specific to a differentiation process. First, a collection of network inference methods are used to predict the regulatory relationships in individual public datasets. Then, the inferred directional relationships are weighted and integrated together by evaluating against the cell lineage-specific dataset. To test the accuracy of this algorithm, we collected a time-course RNA-Seq dataset during human erythropoiesis to infer regulatory relationships specific to this differentiation process. The resulting erythroid-specific regulatory network reveals novel regulatory relationships activated in erythropoiesis, which were further validated by genome-wide TR4 binding studies using ChIP-seq. These erythropoiesis-specific regulatory relationships were not identifiable by single dataset-based methods or context-independent integrations. Analysis of the predicted targets reveals that they are all closely associated with hematopoietic lineage differentiation.

Modeling dynamic functional relationship networks and application to ex vivo human erythroid differentiation

MOTIVATION

Functional relationship networks, which summarize the probability of co-functionality between any two genes in the genome, could complement the reductionist focus of modern biology for understanding diverse biological processes in an organism. One major limitation of the current networks is that they are static, while one might expect functional relationships to consistently reprogram during the differentiation of a cell lineage. To address this potential limitation, we developed a novel algorithm that leverages both differentiation stage-specific expression data and large-scale heterogeneous functional genomic data to model such dynamic changes. We then applied this algorithm to the time-course RNA-Seq data we collected for ex vivo human erythroid cell differentiation.

RESULTS

Through computational cross-validation and literature validation, we show that the resulting networks correctly predict the (de)-activated functional connections between genes during erythropoiesis. We identified known critical genes, such as HBD and GATA1, and functional connections during erythropoiesis using these dynamic networks, while the traditional static network was not able to provide such information. Furthermore, by comparing the static and the dynamic networks, we identified novel genes (such as OSBP2 and PDZK1IP1) that are potential drivers of erythroid cell differentiation. This novel method of modeling dynamic networks is applicable to other differentiation processes where time-course genome-scale expression data are available, and should assist in generating greater understanding of the functional dynamics at play across the genome during development.

AVAILABILITY AND IMPLEMENTATION

The network described in this article is available at http://guanlab.ccmb.med.umich.edu/stageSpecificNetwork.