FunSimMat update: new features for exploring functional similarity

Biomedical data, computational methods and tools for evaluating disease-disease associations

In recent decades, exploring potential relationships between diseases has been an active research field. With the rapid accumulation of disease-related biomedical data, a lot of computational methods and tools/platforms have been developed to reveal intrinsic relationship between diseases, which can provide useful insights to the study of complex diseases, e.g. understanding molecular mechanisms of diseases and discovering new treatment of diseases. Human complex diseases involve both external phenotypic abnormalities and complex internal molecular mechanisms in organisms. Computational methods with different types of biomedical data from phenotype to genotype can evaluate disease-disease associations at different levels, providing a comprehensive perspective for understanding diseases. In this review, available biomedical data and databases for evaluating disease-disease associations are first summarized. Then, existing computational methods for disease-disease associations are reviewed and classified into five groups in terms of the usages of biomedical data, including disease semantic-based, phenotype-based, function-based, representation learning-based and text mining-based methods. Further, we summarize software tools/platforms for computation and analysis of disease-disease associations. Finally, we give a discussion and summary on the research of disease-disease associations. This review provides a systematic overview for current disease association research, which could promote the development and applications of computational methods and tools/platforms for disease-disease associations.

HESML: a real-time semantic measures library for the biomedical domain with a reproducible survey

BACKGROUND

Ontology-based semantic similarity measures based on SNOMED-CT, MeSH, and Gene Ontology are being extensively used in many applications in biomedical text mining and genomics respectively, which has encouraged the development of semantic measures libraries based on the aforementioned ontologies. However, current state-of-the-art semantic measures libraries have some performance and scalability drawbacks derived from their ontology representations based on relational databases, or naive in-memory graph representations. Likewise, a recent reproducible survey on word similarity shows that one hybrid IC-based measure which integrates a shortest-path computation sets the state of the art in the family of ontology-based semantic measures. However, the lack of an efficient shortest-path algorithm for their real-time computation prevents both their practical use in any application and the use of any other path-based semantic similarity measure.

RESULTS

To bridge the two aforementioned gaps, this work introduces for the first time an updated version of the HESML Java software library especially designed for the biomedical domain, which implements the most efficient and scalable ontology representation reported in the literature, together with a new method for the approximation of the Dijkstra's algorithm for taxonomies, called Ancestors-based Shortest-Path Length (AncSPL), which allows the real-time computation of any path-based semantic similarity measure.

CONCLUSIONS

We introduce a set of reproducible benchmarks showing that HESML outperforms by several orders of magnitude the current state-of-the-art libraries in the three aforementioned biomedical ontologies, as well as the real-time performance and approximation quality of the new AncSPL shortest-path algorithm. Likewise, we show that AncSPL linearly scales regarding the dimension of the common ancestor subgraph regardless of the ontology size. Path-based measures based on the new AncSPL algorithm are up to six orders of magnitude faster than their exact implementation in large ontologies like SNOMED-CT and GO. Finally, we provide a detailed reproducibility protocol and dataset as supplementary material to allow the exact replication of all our experiments and results.

Semantic similarity and machine learning with ontologies

Ontologies have long been employed in the life sciences to formally represent and reason over domain knowledge and they are employed in almost every major biological database. Recently, ontologies are increasingly being used to provide background knowledge in similarity-based analysis and machine learning models. The methods employed to combine ontologies and machine learning are still novel and actively being developed. We provide an overview over the methods that use ontologies to compute similarity and incorporate them in machine learning methods; in particular, we outline how semantic similarity measures and ontology embeddings can exploit the background knowledge in ontologies and how ontologies can provide constraints that improve machine learning models. The methods and experiments we describe are available as a set of executable notebooks, and we also provide a set of slides and additional resources at https://github.com/bio-ontology-research-group/machine-learning-with-ontologies.

Predicting candidate genes from phenotypes, functions and anatomical site of expression

Motivation

Over the past years, many computational methods have been developed to incorporate information about phenotypes for disease–gene prioritization task. These methods generally compute the similarity between a patient’s phenotypes and a database of gene-phenotype to find the most phenotypically similar match. The main limitation in these methods is their reliance on knowledge about phenotypes associated with particular genes, which is not complete in humans as well as in many model organisms, such as the mouse and fish. Information about functions of gene products and anatomical site of gene expression is available for more genes and can also be related to phenotypes through ontologies and machine-learning models.

Results

We developed a novel graph-based machine-learning method for biomedical ontologies, which is able to exploit axioms in ontologies and other graph-structured data. Using our machine-learning method, we embed genes based on their associated phenotypes, functions of the gene products and anatomical location of gene expression. We then develop a machine-learning model to predict gene–disease associations based on the associations between genes and multiple biomedical ontologies, and this model significantly improves over state-of-the-art methods. Furthermore, we extend phenotype-based gene prioritization methods significantly to all genes, which are associated with phenotypes, functions or site of expression.

Availability and implementation

Software and data are available at https://github.com/bio-ontology-research-group/DL2Vec.

Supplementary information

Supplementary data are available at Bioinformatics online.

Inferring ontology graph structures using OWL reasoning

BACKGROUND

Ontologies are representations of a conceptualization of a domain. Traditionally, ontologies in biology were represented as directed acyclic graphs (DAG) which represent the backbone taxonomy and additional relations between classes. These graphs are widely exploited for data analysis in the form of ontology enrichment or computation of semantic similarity. More recently, ontologies are developed in a formal language such as the Web Ontology Language (OWL) and consist of a set of axioms through which classes are defined or constrained. While the taxonomy of an ontology can be inferred directly from the axioms of an ontology as one of the standard OWL reasoning tasks, creating general graph structures from OWL ontologies that exploit the ontologies' semantic content remains a challenge.

RESULTS

We developed a method to transform ontologies into graphs using an automated reasoner while taking into account all relations between classes. Searching for (existential) patterns in the deductive closure of ontologies, we can identify relations between classes that are implied but not asserted and generate graph structures that encode for a large part of the ontologies' semantic content. We demonstrate the advantages of our method by applying it to inference of protein-protein interactions through semantic similarity over the Gene Ontology and demonstrate that performance is increased when graph structures are inferred using deductive inference according to our method. Our software and experiment results are available at http://github.com/bio-ontology-research-group/Onto2Graph .

CONCLUSIONS

Onto2Graph is a method to generate graph structures from OWL ontologies using automated reasoning. The resulting graphs can be used for improved ontology visualization and ontology-based data analysis.

Gene Ontology semantic similarity tools: survey on features and challenges for biological knowledge discovery

Gene Ontology (GO) semantic similarity tools enable retrieval of semantic similarity scores, which incorporate biological knowledge embedded in the GO structure for comparing or classifying different proteins or list of proteins based on their GO annotations. This facilitates a better understanding of biological phenomena underlying the corresponding experiment and enables the identification of processes pertinent to different biological conditions. Currently, about 14 tools are available, which may play an important role in improving protein analyses at the functional level using different GO semantic similarity measures. Here we survey these tools to provide a comprehensive view of the challenges and advances made in this area to avoid redundant effort in developing features that already exist, or implementing ideas already proven to be obsolete in the context of GO. This helps researchers, tool developers, as well as end users, understand the underlying semantic similarity measures implemented through knowledge of pertinent features of, and issues related to, a particular tool. This should empower users to make appropriate choices for their biological applications and ensure effective knowledge discovery based on GO annotations.

InteGO2: a web tool for measuring and visualizing gene semantic similarities using Gene Ontology

BACKGROUND

The Gene Ontology (GO) has been used in high-throughput omics research as a major bioinformatics resource. The hierarchical structure of GO provides users a convenient platform for biological information abstraction and hypothesis testing. Computational methods have been developed to identify functionally similar genes. However, none of the existing measurements take into account all the rich information in GO. Similarly, using these existing methods, web-based applications have been constructed to compute gene functional similarities, and to provide pure text-based outputs. Without a graphical visualization interface, it is difficult for result interpretation.

RESULTS

We present InteGO2, a web tool that allows researchers to calculate the GO-based gene semantic similarities using seven widely used GO-based similarity measurements. Also, we provide an integrative measurement that synergistically integrates all the individual measurements to improve the overall performance. Using HTML5 and cytoscape.js, we provide a graphical interface in InteGO2 to visualize the resulting gene functional association networks.

CONCLUSIONS

InteGO2 is an easy-to-use HTML5 based web tool. With it, researchers can measure gene or gene product functional similarity conveniently, and visualize the network of functional interactions in a graphical interface. InteGO2 can be accessed via http://mlg.hit.edu.cn:8089/ .

The role of ontologies in biological and biomedical research: a functional perspective

Ontologies are widely used in biological and biomedical research. Their success lies in their combination of four main features present in almost all ontologies: provision of standard identifiers for classes and relations that represent the phenomena within a domain; provision of a vocabulary for a domain; provision of metadata that describes the intended meaning of the classes and relations in ontologies; and the provision of machine-readable axioms and definitions that enable computational access to some aspects of the meaning of classes and relations. While each of these features enables applications that facilitate data integration, data access and analysis, a great potential lies in the possibility of combining these four features to support integrative analysis and interpretation of multimodal data. Here, we provide a functional perspective on ontologies in biology and biomedicine, focusing on what ontologies can do and describing how they can be used in support of integrative research. We also outline perspectives for using ontologies in data-driven science, in particular their application in structured data mining and machine learning applications.

Abundant genetic overlap between blood lipids and immune-mediated diseases indicates shared molecular genetic mechanisms

Epidemiological studies suggest a relationship between blood lipids and immune-mediated diseases, but the nature of these associations is not well understood. We used genome-wide association studies (GWAS) to investigate shared single nucleotide polymorphisms (SNPs) between blood lipids and immune-mediated diseases. We analyzed data from GWAS (n~200,000 individuals), applying new False Discovery Rate (FDR) methods, to investigate genetic overlap between blood lipid levels [triglycerides (TG), low density lipoproteins (LDL), high density lipoproteins (HDL)] and a selection of archetypal immune-mediated diseases (Crohn’s disease, ulcerative colitis, rheumatoid arthritis, type 1 diabetes, celiac disease, psoriasis and sarcoidosis). We found significant polygenic pleiotropy between the blood lipids and all the investigated immune-mediated diseases. We discovered several shared risk loci between the immune-mediated diseases and TG (n = 88), LDL (n = 87) and HDL (n = 52). Three-way analyses differentiated the pattern of pleiotropy among the immune-mediated diseases. The new pleiotropic loci increased the number of functional gene network nodes representing blood lipid loci by 40%. Pathway analyses implicated several novel shared mechanisms for immune pathogenesis and lipid biology, including glycosphingolipid synthesis (e.g. FUT2) and intestinal host-microbe interactions (e.g. ATG16L1). We demonstrate a shared genetic basis for blood lipids and immune-mediated diseases independent of environmental factors. Our findings provide novel mechanistic insights into dyslipidemia and immune-mediated diseases and may have implications for therapeutic trials involving lipid-lowering and anti-inflammatory agents.

FSim: a novel functional similarity search algorithm and tool for discovering functionally related gene products

BACKGROUND

During the analysis of genomics data, it is often required to quantify the functional similarity of genes and their products based on the annotation information from gene ontology (GO) with hierarchical structure. A flexible and user-friendly way to estimate the functional similarity of genes utilizing GO annotation is therefore highly desired.

RESULTS

We proposed a novel algorithm using a level coefficient-weighted model to measure the functional similarity of gene products based on multiple ontologies of hierarchical GO annotations. The performance of our algorithm was evaluated and found to be superior to the other tested methods. We implemented the proposed algorithm in a software package, FSim, based on R statistical and computing environment. It can be used to discover functionally related genes for a given gene, group of genes, or set of function terms.

CONCLUSIONS

FSim is a flexible tool to analyze functional gene groups based on the GO annotation databases.

Drug-target interaction prediction by integrating chemical, genomic, functional and pharmacological data

In silico prediction of unknown drug-target interactions (DTIs) has become a popular tool for drug repositioning and drug development. A key challenge in DTI prediction lies in integrating multiple types of data for accurate DTI prediction. Although recent studies have demonstrated that genomic, chemical and pharmacological data can provide reliable information for DTI prediction, it remains unclear whether functional information on proteins can also contribute to this task. Little work has been developed to combine such information with other data to identify new interactions between drugs and targets. In this paper, we introduce functional data into DTI prediction and construct biological space for targets using the functional similarity measure. We present a probabilistic graphical model, called conditional random field (CRF), to systematically integrate genomic, chemical, functional and pharmacological data plus the topology of DTI networks into a unified framework to predict missing DTIs. Tests on two benchmark datasets show that our method can achieve excellent prediction performance with the area under the precision-recall curve (AUPR) up to 94.9. These results demonstrate that our CRF model can successfully exploit heterogeneous data to capture the latent correlations of DTIs, and thus will be practically useful for drug repositioning. Supplementary Material is available at http://iiis.tsinghua.edu.cn/~compbio/papers/psb2014/psb2014_sm.pdf.

Analysis of quantitative trait loci in mice suggests a role of Enoph1 in stress reactivity

Significant progress in elucidating the genetic etiology of anxiety and depression has been made during the last decade through a combination of human and animal studies. In this study, we aimed to discover genetic loci linked with anxiety as well as depression in order to reveal new candidate genes. Therefore, we initially tested the behavioral sensitivity of 543 F2 animals derived from an intercross of C57BL/6J and C3H/HeJ mice in paradigms for anxiety and depression. Next, all animals were genotyped with 269 microsatellite markers with a mean distance of 5.56 cM. Finally, a Quantitative Trait Loci (QTL) analysis was carried out, followed by selection of candidate genes. The QTL analysis revealed several new QTL on chromosome 5 with a common core interval of 19 Mb. We further narrowed this interval by comparative genomics to a region of 15 Mb. A database search and gene prioritization revealed Enoph1 as the most significant candidate gene on the prioritization list for anxiety and also for depression fulfilling our selection criteria. The Enoph1 gene, which is involved in polyamine biosynthesis, is differently expressed in parental strains, which have different brain spermidine levels and show distinct anxiety and depression-related phenotype. Our result suggests a significant role in polyamines in anxiety and depression-related behaviors.

Profiling of Parkin-binding partners using tandem affinity purification

Parkinson's disease (PD) is a progressive neurodegenerative disorder affecting approximately 1-2% of the general population over age 60. It is characterized by a rather selective loss of dopaminergic neurons in the substantia nigra and the presence of α-synuclein-enriched Lewy body inclusions. Mutations in the Parkin gene (PARK2) are the major cause of autosomal recessive early-onset parkinsonism. The Parkin protein is an E3 ubiquitin ligase with various cellular functions, including the induction of mitophagy upon mitochondrial depolarizaton, but the full repertoire of Parkin-binding proteins remains poorly defined. Here we employed tandem affinity purification interaction screens with subsequent mass spectrometry to profile binding partners of Parkin. Using this approach for two different cell types (HEK293T and SH-SY5Y neuronal cells), we identified a total of 203 candidate Parkin-binding proteins. For the candidate proteins and the proteins known to cause heritable forms of parkinsonism, protein-protein interaction data were derived from public databases, and the associated biological processes and pathways were analyzed and compared. Functional similarity between the candidates and the proteins involved in monogenic parkinsonism was investigated, and additional confirmatory evidence was obtained using published genetic interaction data from Drosophila melanogaster. Based on the results of the different analyses, a prioritization score was assigned to each candidate Parkin-binding protein. Two of the top ranking candidates were tested by co-immunoprecipitation, and interaction to Parkin was confirmed for one of them. New candidates for involvement in cell death processes, protein folding, the fission/fusion machinery, and the mitophagy pathway were identified, which provide a resource for further elucidating Parkin function.

Inferring potential microRNA-microRNA associations based on targeting propensity and connectivity in the context of protein interaction network

MicroRNAs (miRNAs) are a group of small non-coding RNAs that play important regulatory roles at the post-transcriptional level. Although several computational methods have been developed to compare miRNAs, it is still a challenging and a badly needed task with the availability of various biological data resources. In this study, we proposed a novel graph theoretic property based computational framework and method, called miRFunSim, for quantifying the associations between miRNAs based on miRNAs targeting propensity and proteins connectivity in the integrated protein-protein interaction network. To evaluate the performance of our method, we applied the miRFunSim method to compute functional similarity scores of miRNA pairs between 100 miRNAs whose target genes have been experimentally supported and found that the functional similarity scores of miRNAs in the same family or in the same cluster are significantly higher compared with other miRNAs which are consistent with prior knowledge. Further validation analysis on experimentally verified miRNA-disease associations suggested that miRFunSim can effectively recover the known miRNA pairs associated with the same disease and achieve a higher AUC of 83.1%. In comparison with similar methods, our miRFunSim method can achieve more effective and more reliable performance for measuring the associations of miRNAs. We also conducted the case study examining liver cancer based on our method, and succeeded in uncovering the candidate liver cancer related miRNAs such as miR-34 which also has been proven in the latest study.

Dense genotyping of immune-related disease regions identifies nine new risk loci for primary sclerosing cholangitis

Primary sclerosing cholangitis (PSC) is a severe liver disease of unknown etiology leading to fibrotic destruction of the bile ducts and ultimately to the need for liver transplantation. We compared 3,789 PSC cases of European ancestry to 25,079 population controls across 130,422 SNPs genotyped using the Immunochip. We identified 12 genome-wide significant associations outside the human leukocyte antigen (HLA) complex, 9 of which were new, increasing the number of known PSC risk loci to 16. Despite comorbidity with inflammatory bowel disease (IBD) in 72% of the cases, 6 of the 12 loci showed significantly stronger association with PSC than with IBD, suggesting overlapping yet distinct genetic architectures for these two diseases. We incorporated association statistics from 7 diseases clinically occurring with PSC in the analysis and found suggestive evidence for 33 additional pleiotropic PSC risk loci. Together with network analyses, these findings add to the genetic risk map of PSC and expand on the relationship between PSC and other immune-mediated diseases.

Chapter 15: disease gene prioritization

Disease-causing aberrations in the normal function of a gene define that gene as a disease gene. Proving a causal link between a gene and a disease experimentally is expensive and time-consuming. Comprehensive prioritization of candidate genes prior to experimental testing drastically reduces the associated costs. Computational gene prioritization is based on various pieces of correlative evidence that associate each gene with the given disease and suggest possible causal links. A fair amount of this evidence comes from high-throughput experimentation. Thus, well-developed methods are necessary to reliably deal with the quantity of information at hand. Existing gene prioritization techniques already significantly improve the outcomes of targeted experimental studies. Faster and more reliable techniques that account for novel data types are necessary for the development of new diagnostics, treatments, and cure for many diseases.

Specificity of linear motifs that bind to a common mitogen-activated protein kinase docking groove

Mitogen-activated protein kinases (MAPKs) have a docking groove that interacts with linear "docking" motifs in binding partners. To determine the structural basis of binding specificity between MAPKs and docking motifs, we quantitatively analyzed the ability of 15 docking motifs from diverse MAPK partners to bind to c-Jun amino-terminal kinase 1 (JNK1), p38α, and extracellular signal-regulated kinase 2 (ERK2). Classical docking motifs mediated highly specific binding only to JNK1, and only those motifs with a sequence pattern distinct from the classical MAPK binding docking motif consensus differentiated between the topographically similar docking grooves of ERK and p38α. Crystal structures of four complexes of MAPKs with docking peptides, representing JNK-specific, ERK-specific, or ERK- and p38-selective binding modes, revealed that the regions located between consensus positions in the docking motifs showed conformational diversity. Although the consensus positions in the docking motifs served as anchor points that bound to common MAPK surface features and mostly contributed to docking in a nondiscriminatory fashion, the conformation of the intervening region between the anchor points mostly determined specificity. We designed peptides with tailored MAPK binding profiles by rationally changing the length and amino acid composition of intervening regions located between anchor points. These results suggest a coherent structural model for MAPK docking specificity that reveals how short linear motifs binding to a common kinase docking groove can mediate diverse interaction patterns and contribute to correct MAPK partner selection in signaling networks.

Mining GO annotations for improving annotation consistency

Despite the structure and objectivity provided by the Gene Ontology (GO), the annotation of proteins is a complex task that is subject to errors and inconsistencies. Electronically inferred annotations in particular are widely considered unreliable. However, given that manual curation of all GO annotations is unfeasible, it is imperative to improve the quality of electronically inferred annotations. In this work, we analyze the full GO molecular function annotation of UniProtKB proteins, and discuss some of the issues that affect their quality, focusing particularly on the lack of annotation consistency. Based on our analysis, we estimate that 64% of the UniProtKB proteins are incompletely annotated, and that inconsistent annotations affect 83% of the protein functions and at least 23% of the proteins. Additionally, we present and evaluate a data mining algorithm, based on the association rule learning methodology, for identifying implicit relationships between molecular function terms. The goal of this algorithm is to assist GO curators in updating GO and correcting and preventing inconsistent annotations. Our algorithm predicted 501 relationships with an estimated precision of 94%, whereas the basic association rule learning methodology predicted 12,352 relationships with a precision below 9%.

Semantic similarity analysis of protein data: assessment with biological features and issues

The integration of proteomics data with biological knowledge is a recent trend in bioinformatics. A lot of biological information is available and is spread on different sources and encoded in different ontologies (e.g. Gene Ontology). Annotating existing protein data with biological information may enable the use (and the development) of algorithms that use biological ontologies as framework to mine annotated data. Recently many methodologies and algorithms that use ontologies to extract knowledge from data, as well as to analyse ontologies themselves have been proposed and applied to other fields. Conversely, the use of such annotations for the analysis of protein data is a relatively novel research area that is currently becoming more and more central in research. Existing approaches span from the definition of the similarity among genes and proteins on the basis of the annotating terms, to the definition of novel algorithms that use such similarities for mining protein data on a proteome-wide scale. This work, after the definition of main concept of such analysis, presents a systematic discussion and comparison of main approaches. Finally, remaining challenges, as well as possible future directions of research are presented.

GOrevenge: a novel generic reverse engineering method for the identification of critical molecular players, through the use of ontologies

The ever-increasing use of ontologies in modern biological analysis and interpretation facilitates the understanding of the cellular procedures, their hierarchical organization, and their potential interactions at a system's level. Currently, the gene ontology serves as a paradigm, where through the annotation of whole genomes of certain organisms, genes subsets selected, either from high-throughput experiments or with an established pivotal role regarding the probed disease, can act as a starting point for the exploration of their underlying functional interconnections. This may also aid the elucidation of hidden regulatory mechanisms among genes. Reverse engineering the functional relevance of genes to specific cellular pathways and vice versa, through the exploitation of the inner structure of the ontological vocabularies, may help impart insight regarding the identification and prioritization of the critical role of specific genes. The proposed graph-theoretical method is showcased in a pancreatic cancer and a T-cell acute lymphoblastic leukemia gene set, incorporating edge and Resnik semantic similarity metrics, and systematically evaluated regarding its performance.

Improving disease gene prioritization using the semantic similarity of Gene Ontology terms

MOTIVATION

Many hereditary human diseases are polygenic, resulting from sequence alterations in multiple genes. Genomic linkage and association studies are commonly performed for identifying disease-related genes. Such studies often yield lists of up to several hundred candidate genes, which have to be prioritized and validated further. Recent studies discovered that genes involved in phenotypically similar diseases are often functionally related on the molecular level.

RESULTS

Here, we introduce MedSim, a novel approach for ranking candidate genes for a particular disease based on functional comparisons involving the Gene Ontology. MedSim uses functional annotations of known disease genes for assessing the similarity of diseases as well as the disease relevance of candidate genes. We benchmarked our approach with genes known to be involved in 99 diseases taken from the OMIM database. Using artificial quantitative trait loci, MedSim achieved excellent performance with an area under the ROC curve of up to 0.90 and a sensitivity of over 70% at 90% specificity when classifying gene products according to their disease relatedness. This performance is comparable or even superior to related methods in the field, albeit using less and thus more easily accessible information.

AVAILABILITY

MedSim is offered as part of our FunSimMat web service (http://www.funsimmat.de).