Marzieh Ayati

MOBAS: identification of disease-associated protein subnetworks using modularity-based scoring

Network-based analyses are commonly used as powerful tools to interpret the findings of genome-wide association studies (GWAS) in a functional context. In particular, identification of disease-associated functional modules, i.e., highly connected protein-protein interaction (PPI) subnetworks with high aggregate disease association, are shown to be promising in uncovering the functional relationships among genes and proteins associated with diseases. An important issue in this regard is the scoring of subnetworks by integrating two quantities: disease association of individual gene products and network connectivity among proteins. Current scoring schemes either disregard the level of connectivity and focus on the aggregate disease association of connected proteins or use a linear combination of these two quantities. However, such scoring schemes may produce arbitrarily large subnetworks which are often not statistically significant or require tuning of parameters that are used to weigh the contributions of network connectivity and disease association.Here, we propose a parameter-free scoring scheme that aims to score subnetworks by assessing the disease association of interactions between pairs of gene products. We also incorporate the statistical significance of network connectivity and disease association into the scoring function. We test the proposed scoring scheme on a GWAS dataset for two complex diseases type II diabetes (T2D) and psoriasis (PS). Our results suggest that subnetworks identified by commonly used methods may fail tests of statistical significance after correction for multiple hypothesis testing. In contrast, the proposed scoring scheme yields highly significant subnetworks, which contain biologically relevant proteins that cannot be identified by analysis of genome-wide association data alone. We also show that the proposed scoring scheme identifies subnetworks that are reproducible across different cohorts, and it can robustly recover relevant subnetworks at lower sampling rates.

Prioritization of genomic locus pairs for testing epistasis

In recent years, genome-wide association studies (GWAS) have successfully identified loci that harbor genetic variants associated with complex diseases. However, susceptibility loci identified by GWAS so far generally account for a limited fraction of heritability in patient populations. More recently, there has been considerable attention on identifying epistatic interactions. However, the large number of pairs to be tested for epistasis poses significant challenges, in terms of both computational (run-time) and statistical (multiple hypothesis testing) considerations. In this paper, we propose a new method to reduce the number of tests required to identify epistatic pairs of genomic loci. The key idea of the proposed algorithm is to reduce the data by identifying sets of loci that may be complementary in their association with the disease. Namely, we identify population covering locus sets (PoCos), i.e., sets of loci that harbor at least one susceptibility allele in samples with the phenotype of interest. Then we compute representative genotypes for PoCos, and assess the significance of the interactions between pairs of PoCos. We use the results of this assessment to prioritize pairs of loci to be tested for epistasis. We test the proposed method on two independent GWAS data sets of Type 2 Diabetes (T2D). Our experimental results show that the proposed method reduces the number of hypotheses to be tested drastically, enabling efficient identification of more epistatic loci that are statistically significant. Moreover, some of the identified epistatic pairs of loci are reproducible between the two datasets. We also show that the proposed method outperforms an existing method for prioritization of locus pairs.

Overcoming drug resistance by co-targeting

Removal or suppression of key proteins in an essential pathway of a pathogen is expected to disrupt the pathway and prohibit the pathogen from performing a vital function. Thus disconnecting multiple essential pathways should disrupt the survival of a pathogen even when it has multiple pathways to drug resistance. We consider a scenario where the drug-resistance pathways are unknown. To disrupt these pathways, we consider a cut set S of G, where G is a connected simple graph representing the protein interaction network of the pathogen, so that G-S splits to two partitions such that the endpoints of each pathway are in different partitions. If the difference between the sizes of the two partitions is high, the probability of existence of a functioning pathway in one partition is increased. Thus, we need to partition the graph into two balanced partitions. We approximate the balanced bipartitioning problem with spectral bipartitioning since finding (2,1)-separator is NP-complete. We test our technique on E. coli and C. jejuni. We show that over 50% of genes in the cut sets are essential. Moreover, all proteins in the cut sets have fundamental roles in cell and inhibition of each of them is harmful for cell survival. Also, 20% and 17% of known targets are in the vertex cut of E. coli and C. jejuni. Hence our approach has produced plausible “co-targets” whose inhibition should counter a pathogens drug resistance.

Phosphoproteomics Profiling of Nonsmall Cell Lung Cancer Cells Treated with a Novel Phosphatase Activator

Activation of protein phosphatase 2A (PP2A) is a promising anticancer therapeutic strategy, as this tumor suppressor has the ability to coordinately downregulate multiple pathways involved in the regulation of cellular growth and proliferation. In order to understand the systems‐level perturbations mediated by PP2A activation, we carried out mass spectrometry‐based phosphoproteomic analysis of two KRAS mutated non‐small cell lung cancer (NSCLC) cell lines (A549 and H358) treated with a novel small molecule activator of PP2A (SMAP). Overall, this permitted quantification of differential signaling across over 1600 phosphoproteins and 3000 phosphosites. Kinase activity assessment and pathway enrichment implicate collective downregulation of RAS and cell cycle kinases in the case of both cell lines upon PP2A activation. However, the effects on RAS‐related signaling are attenuated for A549 compared to H358, while the effects on cell cycle‐related kinases are noticeably more prominent in A549. Network‐based analyses and validation experiments confirm these detailed differences in signaling. These studies reveal the power of phosphoproteomics studies, coupled to computational systems biology, to elucidate global patterns of phosphatase activation and understand the variations in response to PP2A activation across genetically similar NSCLC cell lines.

PoCos: Population Covering Locus Sets for Risk Assessment in Complex Diseases

Susceptibility loci identified by GWAS generally account for a limited fraction of heritability. Predictive models based on identified loci also have modest success in risk assessment and therefore are of limited practical use. Many methods have been developed to overcome these limitations by incorporating prior biological knowledge. However, most of the information utilized by these methods is at the level of genes, limiting analyses to variants that are in or proximate to coding regions. We propose a new method that integrates protein protein interaction (PPI) as well as expression quantitative trait loci (eQTL) data to identify sets of functionally related loci that are collectively associated with a trait of interest. We call such sets of loci “population covering locus sets” (PoCos). The contributions of the proposed approach are three-fold: 1) We consider all possible genotype models for each locus, thereby enabling identification of combinatorial relationships between multiple loci. 2) We develop a framework for the integration of PPI and eQTL into a heterogenous network model, enabling efficient identification of functionally related variants that are associated with the disease. 3) We develop a novel method to integrate the genotypes of multiple loci in a PoCo into a representative genotype to be used in risk assessment. We test the proposed framework in the context of risk assessment for seven complex diseases, type 1 diabetes (T1D), type 2 diabetes (T2D), psoriasis (PS), bipolar disorder (BD), coronary artery disease (CAD), hypertension (HT), and multiple sclerosis (MS). Our results show that the proposed method significantly outperforms individual variant based risk assessment models as well as the state-of-the-art polygenic score. We also show that incorporation of eQTL data improves the performance of identified POCOs in risk assessment. We also assess the biological relevance of PoCos for three diseases that have similar biological mechanisms and identify novel candidate genes. The resulting software is publicly available at http://compbio.case.edu/pocos/.

What Do We Learn from Network-Based Analysis of Genome-Wide Association Data?

Network based analyses are commonly used as powerful tools to interpret the findings of genome-wide association studies (GWAS) in a functional context. In particular, identification of disease-associated functional modules, i.e., highly connected protein-protein interaction (PPI) subnetworks with high aggregate disease association, are shown to be promising in uncovering the functional relationships among genes and proteins associated with diseases. An important issue in this regard is the scoring of subnetworks by integrating two quantities that are not readily compatible: disease association of individual gene products and network connectivity among proteins. Current scoring schemes either disregard the level of connectivity and focus on the aggregate disease association of connected proteins or use a linear combination of these two quantities. However, such scoring schemes may produce arbitrarily large subnetworks which are often not statistically significant, or require tuning of parameters that are used to weigh the contributions of network connectivity and disease association. Here, we propose a parameter-free scoring scheme that aims to score subnetworks by assessing the disease association of pairwise interactions and incorporating the statistical significance of network connectivity and disease association. We test the proposed scoring scheme on a GWAS dataset for type II diabetes (T2D). Our results suggest that subnetworks identified by commonly used methods may fail tests of statistical significance after correction for multiple hypothesis testing. In contrast, the proposed scoring scheme yields highly significant subnetworks, which contain biologically relevant proteins that cannot be identified by analysis of genome-wide association data alone.

CoPhosK: A Method for Comprehensive Kinase Substrate Annotation Using Co-phosphorylation Analysis

We present CoPhosK to predict kinase-substrate associations for phosphopeptide substrates detected by mass spectrometry (MS). The tool utilizes a Naïve Bayes framework with priors of known kinase-substrate associations (KSAs). Through the mining of MS data for the collective dynamic signatures of the kinases’ substrates, as revealed by correlation analysis of phosphopeptide intensity data, the tool infers KSAs in the data for the considerable body of substrates lacking such annotations. We benchmarked the tool against existing approaches for predicting KSAs that rely on static information (e.g. sequences, structures and interactions) using publically available MS data, including breast and ovarian cancer models. The benchmarking reveals that co-phosphorylation analysis can improve prediction performance when static information is available (about 35% of sites) while providing reliable predictions for the remainder, tripling the KSAs available from the experimental MS data to comprehensively and reliably characterize the landscape of kinase-substrate interactions well beyond current limitations.

MoBaS on Phosphorylation Data

Although advances in high-throughput omics technologies revolutionized our understanding of the genomic underpinnings of cancer, there are still many challenges in understanding how patients with common driver mutations may display diverging phosphoproteomic responses to the same treatment. Thus, an examination of the signaling landscape will provide essential molecular information for modeling personalized patient treatment design. However, integrative bioinformatics approaches to identify phosphoproteomics-based molecular states are in their infancy. To address this challenge, we integrated several bioinformatics tools to compare and contrast the drug-induced global signaling alterations of two KRAS mutated non-small cell lung cancer (NSCLC) cell lines, A549 and H358, treated with a novel activator of the tumor suppressor Protein Phosphatase 2A (PP2A) versus DMSO control. To identify protein subnetworks that are enriched in differentially phosphorylated proteins, we use our algorithm MoBaS which is designed to identify protein subnetworks that are enriched in disease-associated genomic variants identified by genome-wide association studies (GWAS). MoBaS takes as input a PPI network and a score for each protein indicating the proteins association with the phenotype of interest. It then identifies protein subnetworks that are (i) composed of densely interacting proteins, and (ii) enriched in proteins with high scores. MoBaS also assesses the statistical significance of the identified subnetworks using permutation tests that enable multiple hypothesis testing. Subsequently, we do kinase enrichment within the statistical significant modules and find Aurora KB as a key kinase differentially regulated between the two cell lines in response to our compound. Further corroborating this finding, Aurora KB was downregulated at the protein and mRNA levels with our treatment in A549 but not in H358. Ultimately, our approach models the diverging protein-level drug response across two similar cell lines representing NSCLC patients. Such phosphoproteomic information will potentially inform the optimal therapy regimen for each individual.

Two scenarios for overcoming drug resistance by co-targeting

Removal of proteins on an essential pathway of a pathogen is expected to prohibit the pathogen from performing a vital function. To disrupt these pathways, we consider a cut set S of simple graph G, where G representing the PPI network of the pathogen. After removing S, if the difference of sizes of two partitions is high, the probability of existence of a functioning pathway is increased. We need to partition the graph into balanced partitions and approximate it with spectral bipartitioning. We consider two scenarios: in the first, we do not have any information on drug targets; in second, we consider information on drug targets. Our databases are E. coli and C. jejuni. In the first scenario, 20% and 17% of proteins in cut of E. coli and C. jejuni are drug targets and in the second scenario 53% and 63% of proteins in cut are drug targets respectively.