YouTube Facebook LinkedIn Google+ Twitter Xinginstagram rss  

By James Golden

Feb. 1, 2008 | Part 2 of this article appeared in the previous issue of Bio-IT World.

In my previous two columns, I discussed the application of Open Source Intelligence (OSINT) to improving pharmaceutical IT and business analysis, and outlined some tools and techniques for building a pharmaceutical intelligence capability. In the final column in this series, I discuss how to make intelligence actionable, with a real example. 

Useful intelligence needs to be actionable. What’s the point of searching for something if you don’t do anything with it? It’s not about search, it’s about find — and doing something once you find it. Useful intelligence needs to be relayed to decision makers who can use it to improve the business, make decisions about a marker or compound, or solve a regulatory question.

Once you create a repeatable system for answering your own or management’s questions, you may be asked to create all manner of regular reports. These reports could include information organized by category including:

•            Indications and Warnings identifying potential actions with the goal of providing sufficient warning to preempt or counter their outcome (a bad press release, a competitor’s announcement, the cancelation of a clinical trial, a warning letter from the FDA, etc.). 

•            Current Intelligence including the integration, evaluation, analysis and interpretation of information regarding persons, products, organizations, competitors, and regulatory bodies.

•            Targeted Intelligence including analysis of specific identified groups or persons, company-related assets (physician’s groups, principle investigators, manufacturing facilities, drug delivery systems, etc.), or vulnerabilities for exploitation.  Identifying and monitoring trends in off-label prescriptions is an excellent example.

•            Scientific and Technical Intelligence that directly effect projects in your pipeline.  These intelligence products cover technical capability; scientific knowledge related to drug creation, license, and use of patents and intellectual property; engineering specifications; medical and biological methods and platforms; etc.

One of the most interesting uses of our OSINT system has been around gauging “sentiment.” Traditionally, OSINT analysts have produced intelligence to help the Department of Defense (DoD) understand sentiment — how local language news sources and government officials regard the United States and its Armed Forces. From a life sciences perspective, this approach can be used for the creation and calculation of scientist’s, thought leader’s or organization’s feelings or emotional response as manifested by descriptions in open sources and changes in rhetoric or opinion. 

For example: does the totality of the biomedical literature really think your mechanism of action is a good candidate for drug discovery? Has sentiment for a particular biomarker’s utility changed over time? Is it still a good bet for investment? How do I quantify and report it to management in an understandable and actionable way? We often produce trend reports for sentiment over time that can be viewed through a dashboard portal. Our partners get access to the underlying trend data and quantitative metrics, plus we put some additional analysis on top.

A Pharmaceutical OSINT Tool Kit
So, now we have a pharmaceutical OSINT tool kit. This workbench includes a solid grounding in intelligence philosophy about how to answer questions, and a set of tools and techniques that could include:

1.         OSINT search and content extraction tools (RSS aggregators, Web crawling, and scraping);

2.           Content classification and tagging tools;

3.           Categorization and clustering tools;

4.          Entity and relationship extraction tools;

5.           Taxonomy/Ontology creation and management tools;

6.           Presentation software (including visualization, network descriptions);

7.            Desktop search engines (personal knowledge refinement by individual analysts);

8.            Analytics tools (increased chatter, number of mentions, sentiment, opinion).

Let’s imagine a real-world problem to test this approach. Imagine you’re a member of a product development team around imaging biomarkers. Your company, Global MegaPharm, wants to identify imaging biomarkers that will help validate a therapeutic product in the company’s pipeline and could also be a product in and of itself. Among the burning intelligence questions you might want answered include:

1.         What are the subject-specific keywords for internal searching to see what’s been done to date within our own company?

2.         What are the relevant journal abstracts and papers regarding the putative biomarker?

3.         What are some of the trends around this research, e.g. volume of papers, ranking of papers by citations, author status, gauging of “sentiment?”

4.         Can we identify the source(s) of cutting-edge research from the authors’ affiliations?

5.         Can we rank those institutions and universities by their relevance to our project?

6.         Can we identify thought leaders and departments within those universities?

7.         How can we identify upcoming seminar topics and speakers?

8.         Can we spot trends and predict future developments around our putative biomarker?

9.         With whom are those thought leaders working?

If done right, your OSINT tool kit could help you assemble a workflow for answering these questions quickly and efficiently. Using our pharmaceutical OSINT approach you could:

1.            Identify relevant (if not seminal) academic papers and reviews for understanding the imaging biomarker;

2.         Use automated entity and relationship extraction software to create a keyword list;

3.         Use this keyword list to identify PubMed abstracts and full-text articles for further investigation;

4.         Run entity and relationship extraction software on this new corpus of information;

5.            Review keyword lists for accuracy and completeness;

6.         Use revised keyword lists to pull information from open sources utilizing deep web content mining (journal articles, news sources, meeting abstracts, blogs, government grants);

7.            Expand keyword lists into a taxonomy or ontology for improving enterprise-wide search. Your computational biology group should be overjoyed to receive a well constructed biomarker ontology from the product teams!

8.            Calculate trends including number of papers, thought leaders, top institutions, and key concepts to identify sentiment;

9.            Cluster people, institutions and concepts based on trend numbers and relationships to each other;

10.            Create relationship maps between imaging biomarker concepts, people and places;

11.            Identify additional intelligence and drill deeper into individual parts of the process;

12.            Deliver intelligence to your team on a regular basis that includes updated keywords, OSINT sources, trend metrics, leading industry and academic researchers.

Intelligence is about answering questions. OSINT is about finding information needed to answer those questions in public sources. While search is an important component, it’s not the only component for creating actionable intelligence that will help you improve the business and drive your organization forward.

Jim Golden is a CTO at SAIC. He can be reached at 


This article appeared in Bio-IT World Magazine. 
Subscriptions are free for qualifying individuals. 
 Apply Today.

For reprints and/or copyright permission, please contact Angela Parsons, 781.972.5467.