Drug repositioning using fusion of in silico information sources

OData support
Supervisor:
Dr. Antal Péter
Department of Measurement and Information Systems

The drug repositioning - reuse of registered medicinal substances in new indications -

is becoming popular as the cost of a new drug development process is increasing in the

pharmaceutical industry, whereas the yearly number of new patented drugs is decreasing.

Finding new therapeutic areas of existing and tested drugs is not only of economic interest,

but social too.

The joint study of the side effect profile and chemical structure of drugs allows a radically

different drug similarity metric. It relies on the hypothesis if the side effect profile of

two drugs is similar in several side effects, their targets are probably the same, or at least

they are in the same pathway. In some sense this approach outperforms techniques based

on chemical similarity measure because it measures the effect of the drug on the organism

as whole. Therefore the application of chemical descriptors and side effect similarity in a

common framework is desirable.

In this paper we overview earlier results for the hypothesis of common proteins in case

of similar joint chemical and side effect profiles. Then we propose improvements such as

using publicly available placebo-controlled side effect prevalence data instead of simple term

frequency based text mining approach., and using many heterogeneous systems biological

data sources.

Using large number of heterogeneous data sources requires a new approach which is

based on the combination of these similarities in the kernel space. This idea has already

been successfully applied in gene prioritization using heterogeneous data sources.

I adapted this MKL (Multiple Kernel Learning) based approach for the prioritization of

drugs for a new indication.

The long-term goal of our research is to support drug repositioning with the joint usage

of chemical features and side effects, to gain deeper insight of the side effect as a phe-

nomenon, and the effects that influence them. This work has been carried out within a

joint research of the Bioinformatics Research Group at BUTE Department Of Measure-

ment and Information Systems and the Semmelweis University Department of Organic

Chemistry.

Downloads

Please sign in to download the files of this thesis.