DiseaseNetMiner - A novel tool for mining integrated biological networks of host and pathogen interaction

Modern society is increasingly under threat from a plethora of microscopic fungal pathogens, which cause diseases in agricultural and horticultural crops, and in farmed animals. Many of these diseases cause a significant, detrimental impact on global and local food security. Furthermore, there is a worrying tendency for pathogens to become more aggressive towards their hosts, to cause more disease (increased virulence) and for the anti-fungal chemicals (called fungicides) that are often used to control fungal pathogen outbreaks to become less effective.

In the recent past, (during the genomics era) scientists have developed technologies to sequence and assemble all the chromosomes of an organisms and predict the gene content (i.e. obtain their complete genome blueprints). Now, in the post-genomic era, next generation sequencing technologies have been developed and this has led to an explosion of more genomic data alongside a wealth of gene expression, protein expression, genetic and biological data, which are used by scientists to describe pathogen-host interaction phenotypes and disease outcomes. However, for many scientists with expertise in biology, biochemistry or genetics, this 'omics' data explosion is often seen as a burden, 'an infinite data soup of varying qualities' that only those with specialist computing-based interpretation skills (called bioinformatics), but often only minimal specialist biological knowledge, can penetrate. Therefore, new computer based tools urgently need to be developed to allow researchers to connect, explore and compare all the large and small-scale datasets available for pathogenic species that cause diseases. Once we fully understand how fungal pathogens cause disease, and how the host species try to defend themselves, will it be possible to manipulate these processes and mechanisms and go on to devise new ways to reduce disease levels and thereby improve global food security.

In this project, we will develop a novel software tool, called DiseaseNetMiner, which will be user-friendly and can be used by many different types of scientists to explore integrated biological networks that can predict processes controlling the disease-causing abilities of fungal pathogens. DiseaseNetMiner will deliver understandable outputs from diverse and complex large-scale data inputs. DiseaseNetMiner will allow researchers without specialist bioinformatics skill to explore and compare this wealth of existing data from multiple species with their own latest cutting-edge results to permit rapid progress and new discoveries. This fundamental tool will effectively connect different data types and then return the results in an accessible, explorable, as well as scalable, format that can be easily manipulated, displayed and interrogated. DiseaseNetMiner will create a novel research environment from which new scientific insights and biological discoveries can be made.

The UK research community has been at the very forefront of research and discovery in this field. The initiative we propose will be an exceptionally useful and cost-effective way of ensuring that the leadership shown by the UK research community will continue in the decade ahead. We expect this to yield outcomes with huge impact in our field and beyond, to meet the grand challenges of our age.

DiseaseNetMiner will provide a user-friendly tool for candidate gene prioritisation and hypothesis generation from large host-pathogen knowledge networks. To ensure efficient and cost-effective delivery, we will make use of the free and open-source Ondex and QTLNetMiner frameworks that we have previously developed at Rothamsted Research. DiseaseNetMiner will include significant, new functional advances and provide a timely and novel tool for the plant and fungal research communities.

Objective 1: Use Ondex to integrate public multi-omics datasets, phenotype information, homology data and functional gene annotations for the key fungal model species S. cerevisiae, N. crassa and A. nidulans, and the pathogenic fungi F. graminearum and Z. tritici. This will deliver a semantically integrated knowledge network (graph data warehouse) of interlinked fungal species that we will provide access to through the QTLNetMiner web application (FungiNetMiner).

Objective 2: Create a novel, combined pathogen-host network by integrating an existing wheat-arabidopsis-rice knowledge network and the new fungal network based on annotations to cross-species ontologies, curated pathogen-host interaction databases, text-mining using gene names and disease/phenotype ontologies, and a statistical correlation approach using data from dual long time-series RNA-seq experiments of infected plants. This knowledge network is estimated to have about 750,000 nodes and 4,000,000 edges including all wheat, Arabidopsis and fungal genes.

Objective 3: Develop the DiseaseNetMiner web application by extending the QTLNetMiner client-server framework to fully support combined plant-pathogen networks and user-provided SNP/GWAS input data. This will require the development of new graph queries to explore combined plant-pathogen networks and novel tools to visualise SNPs within biological interaction networks and to include SNP consequences in the QTLNetMiner gene prioritisation algorithm.

Modern society is increasingly under threat from a plethora of fungal pathogens, many of which
have a significant, detrimental impact on food security. In the post-genomic era, the explosion of 'omics, genetic and phenotypic data is often seen as a burden, 'an infinite data soup of varying qualities' that only those with specialist bioinformatics skills, but often only minimal specialist biological knowledge, can penetrate.

In this 15 month pilot project, we will focus on the development of new in silico tools to assist scientists based in academic, industry, NGOs and /or government departments to connect, explore and compare large-scale fungal and plant datasets to identify the various generic and species-specific biological processes that control pathogenesis, fungicide resistance and avirulence, while connecting this information to the general growth and development of eukaryotic organisms. The new tool, called DiseaseNetMiner, will allow researchers without specialist bioinformatics skills to explore and compare this wealth of existing data from multiple species, with their own latest cutting-edge results to permit rapid progress, new insights and new discoveries.

DiseaseNetMiner will contain and connect information in a combined network for three well studied fungal species (S. cerevisiae, N. crassa and A. nidulans) and two globally economically important fungal pathogens of wheat crops, namely Fusarium graminearum and Zymoseptoria tritici. This fungal network will be connected to a plant network combining information from Arabidopsis, rice and wheat.

Academics will use DiseaseNetMiner to interrogate and prioritise candidate gene lists for both pathogens by gaining additional annotation from model species, the responses of the host plant and from the occurrence of SNPs. Researchers would then test the function of the most promising candidate genes with reverse genetics experiments. Researchers will use the combined plant and pathogen networks to explore functions of genes and their interactions. Researchers involved in sequencing large numbers of natural or laboratory generated strains of F. graminearum and Z. tritici with subtly different phenotypes, e.g. increased aggressiveness, can use DiseasesNetMiner for the biological interpretation of genome wide association studies. This should reveal the domains, hubs and /or proteins in the network most likely to be causally linked to the phenotypic shift.

In the short term, the agrochemical and plant breeding industries will use DiseasesNetMiner to connect publically-available datasets on pathogens, crop plants and model species to their own proprietary data sets. This will give new insights on where product failure, (agrochemical or resistant wheat cultivar) may be occurring as individual population shift occur or when individual isolates in a population mutate and rise in abundance. In the longer term, these connected data sets, when viewed over yearly time series, will help to guide the identification of novel target sites for intervention and, therefore, guide innovative agrochemical product development and influence disease resistance breeding strategies, such as identifying the most effective R gene stacks.

NGOs and government departments will use DiseaseNetMiner to investigate and predict the possibility of emerging disease threats in the UK and elsewhere by investigating the genomic changes present in newly virulence strains and/ or fungicide resistant strains and go on to provide advice to farmers and the AgriIndustry.

To deliver the above impacts, we plan to develop a community of researchers in the UK and elsewhere to beta test the tools with their own unpublished data, publicise the existence of DiseaseNetMiner to the wider scientific community at national/ international conferences, hold virtual and face-to-face workshops, give seminars to various industries / consortia and complete various media activities.