Gene-expression connectivity mapping and its application in phenotypic targeting

When a small-molecule compound is applied to a biological system (cells, animal, or human), the system interacts with (or responds to) the chemical; as a result the very large number of genes in the system change their expressions in a particular pattern. Modern biotechnologies allow researchers to measure those changes in gene expression using 'microarray chips', generating gene-expression profiles. The proposed research involves the development and application of methods to connect different biological conditions using those gene-expression profiles, based on a novel concept called 'connectivity map'. The connectivity map concept was introduced in 2006 in a paper published in Science with the idea: 1) different biological states have their own characteristic gene-expression profiles, and 2) connections between different biological states can be established base on their gene-expression similarity or dissimilarity. This is a simple yet very powerful idea. Two immediate applications of the connectivity mapping concept are: 1) if two chemical compounds are found to induce similar gene-expression profiles, the knowledge about one compound may help to predict the properties (pharmacological as well as toxicological) of the other; 2) if a compound is found to induce a gene-expression profile opposite to that of a disease state (whether animal disease or human disease), the compound may then be flagged as a potential drug to treat the disease. In the past few years, studies on connectivity mapping including ours have demonstrated the success and promise of this approach, for example, in identifying pro-oestrogen (agonists) and anti-oestrogen (antagonists) chemicals, immunosuppressive drugs, hair-growth agents, etc. We will further develop the method and technique to effectively and accurately map the gene-drug-disease connections. The results from this project will benefit biomedical researchers in their search for small-molecule therapeutics to diseases.

The project covers three aspects of research in connectivity mapping: 1) Methodological and algorithmic advancement, 2) Novel application, and 3) software development. We will investigate and develop a gene progression method in the construction of gene signatures, first by ranking and sorting genes based on their statistical and biological significance. By progressively incorporating more features, the length of a gene signature is optimised by observing and achieving the desired statistical stringency among the discovered connections. We will develop a gene-signature perturbation method to test the stabilities of discovered connections. Two approaches to gene-signature perturbation will be investigated and their performance compared. The first is a leave-one-gene-out approach to derive shorter signatures. The second is an add-one-gene-in approach by adding an extra randomly selected gene to the signature. The perturbation stability will be assessed by the proportion of persistent significance. This will help further analyse and narrow down these connections in an unbiased manner such that the success rate of any follow-up investigation can be maximised. We will develop two types of applications following 'the general principle of phenotypic targeting'. Using experimentally derived gene signatures, we will investigate phenotypes related cancers in two case studies. Using manually coded gene signatures, we will apply connectivity mapping to the search for agents that will down regulate a small number of selected genes. We will use RAN and c-FLIP down-regulation as case studies. Finally there is a software development component, in which we will implement a parallel computing model of connectivity mapping with the new methods and algorithms developed here. We will use C/C++ and OpenMP as a development platform to create tools that fully exploit the hardware capacities of the increasingly common multi-core desktop computers.

In addition to academic beneficiaries, biotech and pharmaceutical industry would be interested in the research, as connectivity mapping fits well in suggesting new uses for existing drugs. The regulatory agencies may also be interested in the development of connectivity mapping technology for its potential application in predictive toxicology and consequently reducing the use of animal testing. Identifying new uses for existing drugs becomes an approach of growing popularity with drug developers because of the high-cost and time-consuming cycle associated with new drug discovery. However most remarketed old drugs for new uses so far have been the result of chance observation, the connectivity mapping technique can offer a more targeted screening approach in finding new uses. The longer term economical implication of the development and successful applications of connectivity mapping is time and cost saving for drug developers and for the wider economy. The improved methodology and the statistical rigor we introduced into the connectivity mapping exercise led to increased specificity and sensitivity in identifying meaningful biological connections. Our work has generated a great deal of interest from both academia and industry. To increase further impact, we plan to explore the possibility of exploiting commercially our research results. With the professional support from the Knowledge Exploitation Unit (KEU) at QUB, the applicant plans to undertake consultancy activities in the area of gene-expression connectivity mapping, particularly in the identification of new uses for existing drugs, which is closely related to the proposed research and directly benefits from the findings of this project. These activities will provide an important platform to transfer our knowledge and expertise to industry, to the public and wider community.