How to Capture Market Share Through Data-Driven Drug Discovery

How to Capture Market Share Through Data-Driven Drug Discovery

For years pharmaceutical companies have taken existing drugs and repurposed or repositioned them to create new drugs. These new drugs are often geared towards neglected therapeutic areas. Drug repurposing has always been a solid “go to” strategy and a proven method for mitigating the financial risk of drug discovery and a means of addressing softening market share.

However, since the advent of big-data tools, such as machine learning and natural language processing, coupled with low-cost, high-performance computing, mining “legacy” and “open source” drug discovery data for new insights has moved from “stop-gap” to market disruptor. The model is rapidly becoming a standard innovation approach for established pharma and for many biotech startups – it’s their predicated approach (VC firms are pointing the way).

Simply throwing tools and cheap processing power at the drug discovery problem is not a guaranteed operational approach. The types of analytics involved are complex and vary. From interpreting assay outputs to identifying targets across different patient populations, the data sets involved are complex and multidimensional.

Common pitfalls
In our experience, machine learning programs typically fail across some common areas:

  • High rate of false positive results undermine credibility – typically due to a mismatched correlation between the search thesis and pooled data sets.
  • Integrating data back into varied and multiple data stores becomes burdensome and complex – a symptom of search models is that they have too many variables or a fundamental misunderstanding of data entities.
  • Even after potential targets are identified, operationalizing the results is contingent on the organization being able to operationalize data (ETL maturity continues to be a challenge).
  • Lastly, lack of a clear linkage between the project and a specific business goal or issue sets the program up for failure from the start to be nothing more than an interesting exercise.
    Many of these common pitfalls relate to the quality of historical data and the consistency of data use across the enterprise.

Many of these common pitfalls relate to the quality of historical data and the consistency of data use across the enterprise.

Master data management (MDM) best practices
Employing these best practices can help set your organization’s machine learning program up for success:

  • Focus on standardization and nomenclature management to provide consistency in the definition and use of data attributes
  • Assess and classify third party and external data to align to your organization’s nomenclature prior to using it
  • Assess data quality and fix false positives early through short pilots that focus on term matching
  • Classify and enforce business-driven data hierarchies in order to speed up data transformation processing

A well-executed MDM program supports and complements machine learning projects. A few small steps can lead to improved operational speed, quality of results, and help ensure the machine learning program can scale.

Re-evaluate the path forward
We advise pharmaceutical companies exploring an investment in firms that are focused on drug repositioning and repurposing to do their diligence when it comes to their advanced analytics. Strong capabilities are a must if drug repurposing is a major component of the growth thesis. Areas to look for include machine learning, integration capabilities, and ability to decipher data for key insights.

For those firms that have already started or have integrated machine learning into their bioinformatics process, and are now wondering why they are not seeing a return on advanced analytics investment, it may be time for an MDM refresher.

Phone: 312-602-4000
222 W. Adams
Chicago, IL 60606
Show Buttons
Share On Facebook
Share On Twitter
Share on LinkedIn
Hide Buttons