|
|
|
|
|
|
|
|
|
| CSIRO | SOLVE | Issue 10 | FEB 07 | |
|
| |||||||
|
ARTICLE
BIOTECHNOLOGY:
Drug Discovery on the Double By Dr Whitney MacDonald
A mathematical modelling program is expected to accelerate the trip through the drug discovery pipeline, potentially saving millions of dollars a year per drug.In the early stages of drug discovery, scientists often have thousands of compounds to analyse, and are faced with setting up experiment after experiment to characterise each compound one variable at a time – a process that is costly and time-consuming. However, with the use of a new molecular modelling tool, scientists can gather much of this information in less time than it takes to tell someone about the program. The MolSAR™ software program, developed by Dr Dave Winkler and Dr Frank Burden from CSIRO Molecular and Health Technologies, allows scientists to use a limited set of experimental data to make predictions about the physical or biological properties of chemical compounds and cells. These predictions can be made in seconds while sitting at a computer, without the need to perform the expensive and lengthy ‘lab-bench’ experiments often used to steer further progression through the drug discovery pipeline.
Although modelling tools that make Quantitative Structure-Activity Relationship (QSAR) predictions about a compound have been around for decades, these tools have disadvantages that can increase the amount of error between the model and the experimental outcome and make it difficult for non-expert users to develop valid models. For a start, they need highly skilled modellers to ensure model integrity. However, by using a diverse set of data, novel ways of describing molecules and a robust, nonlinear Bayesian neural network, MolSAR™ allows scientists to build models using their own experimental data. “Bayesian statistics find the best balance between simplicity and complexity, generating a model that captures the underlying information but doesn’t capture the noise in the system,” Dr Winkler says. This makes MolSAR™ a robust option for the general cell biologist or medicinal chemist. This feature was one of the factors that led Bio-Rad Laboratories Inc., an international company specialising in products for life science and diagnostics applications, to license the software and integrate it into their compound analytical software platform, KnowItAll®. Although MolSAR™ has existed for almost 10 years, Bio-Rad Laboratories only picked it up in mid-2006, illustrating the typically long delays between development and commercial pick-up. Dr Gregory Banik, general manager of Bio-Rad’s informatics division, says researchers’ ability to use experimental data generated ‘in-house’, to build local models without needing to write program code, offers many advantages over existing QSAR tools. “While we do have another mechanism for putting local models into the KnowItAll® environment, it’s with a separate application,” Dr Banik says. “That’s the real breakthrough with MolSAR™, it is completely integrated within the KnowItAll® environment, making the building and use of those models much, much easier.” Another advantage MolSAR™ offers is protection of compounds that are not yet patented. “Building a model is most useful in the early stages of drug discovery, when a company is trying to prioritise which leads to follow and which compounds to rule out,” Dr Banik says. “The most useful data for building such models is data generated by the company in the early experiments, yet because of its proprietary nature, the company does not want to outsource the modelling or enter it into a global model. The design of MolSAR™ circumvents this problem.” Professor Stephen Livesey, CEO of the Australian Stem Cell Centre (ASCC), explains how they have also recognised the value that this modelling software has to offer. In collaboration with CSIRO, the ASCC is using MolSAR™ to model interactions within a cellular environment, enabling the centre to better understand adult stem cell biology and embryonic stem cell development. Part of the work focuses on changes in gene expression between two different cell states and analysing the data to determine the trends in gene expression. This involves handling extremely large amounts of data. “To be able to model and define groupings of functional dynamics would help us tremendously when dealing with thousands of genes that are changing at one time,” Professor Livesey says. “It is very important to have an analysis method that can be discriminatory with respect to the type or number of genes that are ultimately investigated.” When working with a complex system containing multiple variables, using the conventional scientific ‘strong arm’ method – approaching it one variable at a time in a logical progression – is a cumbersome and expensive way to get the experimental data. “Being able to interact modelling with experimental data to come up with a predicted model can make our decision in terms of experimental design and implementation much more efficient,” Professor Livesey says.
Ultimately, the MolSAR™ software tool has the potential to fast-track the early ‘grunt work’ period of drug design and stem cell research by increasing the overall efficiency of generating data, specifically by improving experimental design and by better handling large amounts of experimental data. “We were very pleased and proud to partner with CSIRO, as they are clearly a prestigious and internationally well-known organisation, and in particular with Dr Winkler, who is well known in the QSAR field,” Dr Banik says.For further information contact: |
 IN THIS ISSUE
|
| Home | About Us | eSubscribe | Links | |
|
|
||
|
|
|
Back to main