Since 1987 - Covering the Fastest Computers in the World and the People Who Run Them

May 21, 2010

Drug Discovery and Development in the Cloud

Bruce Maches

I hope that all of you have found the information in my blog so far to be of use to you out there. I have received a comment or two regarding potential approaches in dealing with public cloud validation and will provide an update on that topic in a future entry.

So far we have covered some basic facts on the life sciences companies and the regulatory environment they must exist in. I also explored some of the validation issues around using infrastructure related cloud technologies. In this entry I will start to guide you through the basic steps of drug discovery and development at a high level and point out areas where cloud computing can be leverage to facilitate the drug R&D process, reduce costs, and speed time to market. There is obviously a very broad topic and there is no way to effectively cover this subject in just one blog entry so this theme will be the core of the next several posts.

The pharmaceutical R&D process is a long and arduous procedure lasting up to 10 years or longer. It is also tremendously expensive with some new medications costing up to a billion to get from concept to market. There are a variety of factors that impact the cost and duration of the development effort. A few examples are: if a drug is a new one or a different indication (use) for an existing one, what therapeutic area it is targeted for (cancer, diabetes, etc), and what pathway or disease mechanism is the drug addressing. Given all of these factors, the drug development process can be very risky with many potential new drugs never making it to market after consuming hundreds of millions of dollars in research and development efforts. Industry figures vary but on average only 1 compound out of 5,000 or more make it from concept/discovery through development to the market place.

In layman’s terms, the drug development process can be broken into the following high-level activities:

  • Understanding the targeted disease mechanism within the body.
  • Finding a compound that will disrupt or modify that mechanism, initial FDA filing (patent clock is now ticking!).
  • Drug formulation, toxicity studies, and animal safety studies.
  • Initial human testing, further development work, dosing studies.
  • Large scale clinical testing for effectiveness and side effects.
  • Applying to the FDA for market approval.
  • Post approval monitoring.

Many of the applications used to support these pieces of the R&D process can be very compute and storage intensive making them great candidates for moving into the cloud. Given the nature of the drug development process, requirements for compute and storage resources can vary widely with huge peaks in demand as individual experiments or protocols are executed. This makes a strong case for cloud computing as the cost and time necessary to acquire and deploy these types of systems is simply prohibitive for many life science firms. Small companies clearly do not have the budget or resources available to provision these resources internally, and larger firms are dealing with on-going budget constraints with their R&D expenditures. Cloud computing is well suited for ‘bursty’ types of applications as the resources required can be provisioned on demand and at a much lower cost, reducing capital and operational expenses. Also, using cloud significantly cuts the time required to provision and qualify these resources, allowing life science companies to bring their products to market more quickly.

For this post I will concentrate on the up front set of activities around discovery and screening of new compounds and provide some examples of the different aspects of life sciences research that would benefit from the use of cloud computing.

A very commonly used technique in the biotech field is genomic sequencing, which is an extremely data and computationally intensive process. The technology involves looking for specific amino acid sequences in proteins or DNA samples. The amount of data generated is immense with many experimental runs producing gigabytes of data. All of this data has to be managed, stored and made available for follow on research and analysis. One of the applications used in this field is a piece of software called Basic Local Alignment Search Tool, or BLAST. This tool compares amino acid patterns in the sample being analyzed to a library of nucleotides looking for matches to certain sequences. This type of application is well suited to run in the cloud utilizing CPU and storage resources and then bring the results back to the researcher for further study. Tools such as Amazon’s Elastic Cloud Computing and Simple Storage Service (S3) are prime examples of offerings in this area. In addition, Amazon, along with other vendors, maintains copies of many of the publicly-available data sets on genomic and sequence data and makes them available to their clients as part of their overall cloud environment.

Another promising advance in the drug research process is what is called ‘in silico’ or virtual screening. The promise of virtual screening is that it will allow researchers to greatly increase the pace of finding new potential compounds while reducing costs for lab work and clinical trials. The screening process involves using an automated tool to test thousands of compounds for specific activity, either the inhibition or stimulation of a biochemical or biological mechanism. Running these tests requires the preparation of large numbers of assay plates each with hundreds or thousands of tiny wells. Compounds to be tested are placed in the wells using a pipette mechanism, processed by robotic labs and the corresponding reaction recorded. Using this high throughput technique allows researchers to screen thousands to millions of compounds but it can be costly and time consuming. Virtual screening provides the ability to model the desired reaction using tools such as the protein docking algorithm EADock greatly reducing the number of compound combinations needed to be tested. Leveraging cloud resources to perform ‘in silico’ testing will also cut costs and speed time to market.

Other potential applications for cloud computing include research areas such as protein docking simulations, data mining, and molecular modeling. I will reserve those areas for a future entry.

IT organizations supporting life science R&D functions should work towards creating a service based model for how they provide the resources required for these computational and storage thirsty applications. By understanding the underlying cost models and providing clear standards on how/when/where cloud infrastructure will be deployed, the IT group can better be able to properly manage and secure cloud based resources.

Continued advances in the field of drug discovery will exponentially increase the amount of data generated during the discovery process. IT organizations or vendors that can supply the needed cloud based infrastructure services in a secure and reliable manner will certainly do well in this space. Cloud computing also provides significant flexibility to the researcher as they are now free to explore avenues of research that would not have been feasible before the advent of cloud computing.

Cloud computing is certainly expanding its footprint in the life sciences community. The speed, efficiency and cost effectiveness have made cloud computing an indispensible tool for researchers, allowing them to focus on the ‘what’ of science and not the ‘how.’ Having the resources to do better research at this phase of the drug development process will also reduce time and expense in the later phases. My next post will expand further on the challenges and opportunities in the discovery phase of the pharmaceutical research process.

I would love to hear from you if you have any questions or comments. Feel free to contact me at

Share This