As a Data Scientist, I often find myself thinking about potential applications of data science that have yet to be fully realised. One area which I feel is ripe for gain is Research & Development (R&D) in the pharmaceutical sector.
Data science has been transformational in many industries over the past few years. For example, there are countless cases where the application of data science has increased marketing effectiveness, enabled organisations to identify crime or even helped non-governmental organisations (NGOs) determine how best to place resources following major catastrophes such as the earthquakes in Nepal. So how could it help R&D for pharmaceutical organisations?
Pharmaceutical R&D is an activity which is often characterized by diminishing returns and a stale pipeline for new product development. These are characteristics which are not unfamiliar to other areas where data science has proved to be extremely successful. In marketing, for example, the more you spend on a given channel often the less effective it becomes.
There is also a huge array of potential data sources that a pharmaceutical R&D department could look to utilise including:
- The R&D process
- Care providers
- Manufacturing and production lines.
The questions that can be answered by effectively querying this data are almost limitless in the R&D space. For example, this data could be used for:
- Patient screening for clinical trials. Data can be utilised from multiple sources to create a single view of the patient. This may include traditional sources such as medical records but could also include social media and other more unusual forms of data to build an enhanced and more accurate view of the patient’s lifestyle. Genetic data is another obvious avenue; ensuring vital areas are matched leading to smaller sample groups, providing more meaningful data and therefore more cost effective trials.
- Real-time monitoring of clinical trials enabling ‘in the moment’ understanding of issues and events as they arise which in turn allows for a rapid reaction and effective management of risk during a trial phase. This can also be combined with predicative modelling techniques to form an early warning system for events before they arise.
- Predictive modeling of biological processes, which may be used to help identify candidates who have a high probability of success, thus reducing unnecessary time and expenditure by allowing R&D teams to focus on areas that are the most likely to succeed.
- Information from the production line can be used in future formulation to reduce manufacturing costs and maintain product quality.
For Data Science to be effective in pharmaceutical R&D, as in many applications, there are a number of factors that will be crucial:
- A clear data strategy needs to be defined which will allow the unification of data across sources. The data should not be excessively onerous to access. Yet, at the same time, there is a need to recognise that a great deal of the data being processed often is highly sensitive.
- Collaboration should be encouraged; traditionally pharmaceutical R&D has taken place with the utmost levels of secrecy. However, success for data science in R&D will depend both on the ability to grow data rich networks and an ability to develop pioneering techniques in order to overcome some of the unique challenges faced by pharmaceutical R&D departments. For both of these reasons, R&D departments must be willing to look beyond their own silo and extend both internally within their organisation and be willing to engage external parties such as CRO’s or Academics.
- Clear goals should be identified from the start and these should be cumulative in nature. It is unrealistic to expect to go from a very limited use of data science to a fully transformed organisation wholly reliant on the output of data science overnight. Expectations need to be realistic and should be focused around deriving real business benefit.
It remains to be seen whether our large pharmaceutical organisations can overcome their internal blockers and make progress towards these goals. If they can, it is evident that data science has the potential to transform R&D.