New Practices of NLP in a Nordic Pharmaceutical
Novo Nordisk and their use of NLP
I recently read an article written by Jane Z. Reed about NLP and pharmaceutical companies.
Her article is more about how the industry is using NLP than any theoretical discussion.
It must be said straight away that her role is not neutral in this regard as she is the Director of Life Sciences at Linguamatics, an IQVIA company. They are selling NLP services.
In that manner this reads much like a pitch — with a case study as a suggestion to offer her services. Having said that, this does not make her writing less interesting. Rather, it is important to consider someone working with the commercial application of NLP.
The goal of Linguamatics is to: “Advance human health with NLP that transforms your text into positive outcomes.”
She argues that the science behind drug development has become more complex.
In that sense drug-development partners such as biotech startups and university technology transfer offices.
She talks of the Novo Nordisk company as a case [I have added numeration]:
- “Novo Nordisk sought to deepen its pipeline of diabetes and obesity drug candidates via collaborations with external partners.
- Novo Nordisk needed to identify potential partnerships before competitors did so.
- To accomplish this objective, Novo Nordisk had to discover a means of consolidating huge volumes of external information to generate a bird’s-eye-view of partnership opportunities as early as possible.”
So, how can NLP be used for ‘intelligence-gathering’ in the pharmaceutical industry?
She argues the greatest challenge is in: “…identifying collaboration opportunities is scrutinizing the voluminous amount of information that may (or in many cases, may not) contain insightful tidbits about potential drug-development partners.”
Established pharmaceutical companies are looking for new information on:
- Drug development.
- Targets and pathways.
- Biotech companies.
- University technology offerings.
- Clinical trials.
To find these structured and unstructured sources must be examined. Reed mentions:
- News reports.
- Patent filings.
- Scientific papers.
- Conference abstracts
→ She says this goes towards an “evidence hub.”
The ‘evidence hub’: is a curated, data-driven landscape of knowledge.
That does not sound very convincing.
However, she talks of a case with Novo Nordisk.
“Novo Nordisk executives realized that NLP could greatly improve the efficiency of the process of identifying collaboration opportunities by automating text mining to uncover valuable information hidden in troves of unstructured data.”
She suggests text mining as a way to tackle this:
“Text mining: is the process of examining large collections of documents to discover new information or help answer specific research questions. NLP-based text mining empowers computers to, in essence, read text by simulating the human ability to understand a natural language, enabling the analysis of unlimited amounts of text-based data without fatigue in a consistent, unbiased manner.”
She describes one process:
- NLP is used in a semi-automated workflow.
- The workflow uses a suite of NLP queries over data streams coming in from news, patents, scientific literature, conference abstracts and more.
- The resulting outputs are curated into summaries, written by information scientists with experience in the respective therapy areas.
- These are provided via InfoDesk as easy-to-consume alerts to the broader Novo Nordisk researchers.
As such, this is a process of intelligence gathering as outline earlier, certainly structured in a manner, to an extent that makes it more than simply buzzwords.
In addition to this they had a goal of ‘empowering’ members of research team to serve as “scouts” for new partnership opportunities.
Two tools were developed to assist researchers in becoming scouts. One was a newsletter and the second a dashboard with up-to-date ‘landscapes’ for each therapeutic area of interest.
An NLP process in Novo Nordisk
Reed then proceeds to describe a process in Novo Nordisk.
- “A news article is published that profiles a biotech startup investigating an obesity drug candidate.
- Using pre-established search criteria, Novo Nordisk’s NLP evidence hub flags the article and publishes it to the “obesity drug candidate” newsletter and dashboard, prompting a “scout” to write a summary.
- The system also surfaces relevant background on the startup, revealing that it recently raised venture-capital funding, is seeking a patent with a novel method of action for the obesity drug candidate, and is scheduled to present new data at an upcoming conference.
- In combination with these, a Novo Nordisk researcher recognizes that a former colleague is employed by the startup and uses LinkedIn to facilitate a meeting between the former colleague and a Novo Nordisk team member who is attending the conference.”
Clearly, this is one application. One way of gathering information.
Soon, according to Reed, they are planning for Novo Nordisk to: “…capture consumer sentiment by adding new data sources, including publicly available social media posts.”
This is a short insight into commercial NLP in the pharmaceutical industry.
What do you think about these practices?
Let me know with a response.
This is #500daysofAI and you are reading article 434. I am writing one new article about or related to artificial intelligence every day for 500 days.
