In times of the new normal, Life Sciences field representatives have to adopt the hybrid role of face-to-face and remote engagements. The more digital and remote channels become prominent the higher the importance of content-rich materials will be. This is why I have invited a former colleague of mine, Bernd Schossmann, for an interview. Bernd, has been passionate about programming Artificial Intelligence (AI) to extract, amongst other activities, deep insights on scientific publications and is partner of TEGRUM, a recently founded company specialized in this area.
Karsten: Thank you, Bernd, for taking the time for this interview. May I ask you first how you have developed this passion for programming AI?
Bernd: I got started with AI quite early, in the 1980s. I developed expert systems and games on the first home computers. At university I worked on the inversion of Laplace transforms, a problem that is today used in AI, specifically in the Netflix recommender system. After a long hiatus from AI, I entered talks with IBM Watson to join their team in 2014.
I did not take the job but in preparing for the interviews I saw how much AI had progressed. It was astounding! I went through an AI curriculum I had thrown together from various sources on the Internet and it took me a few years to get up to speed. Eventually I developed more and more tools that would help me in my job, this is where Pharmaceutical Artificial Intelligence (PAI) got started.
Karsten: How could you explain PAI to someone who is not at all familiar within this field?
Bernd: PAI distills insight from large texts. Much like a very fast editor it reads and recognizes the key themes and concepts in a text - thousands of pages within seconds.
Karsten: How can Life Sciences companies make best use of the power of PAI in their customer engagements?
Bernd: PAI is mostly a text analysis toolset so the applications are quite broad.
One application is market entry with a new brand. Scientific literature in any indication today is massive, often ranging in the 10s of thousands of studies. PAI quickly maps the scientific landscape, shows what the key concerns and topics of the research are and relates it to the data of various therapies in a given indication. A Life Science company gains insights of what really interests the scientific community and where the challenges of the patients are.
Another application is content generation. To keep HCPs interested you have to engage them with interesting and relevant scientific content. PAI can quickly identify key studies and automatically propose content for newsletters or websites. This has recently gained traction as interactions are increasingly digital and content creation is really the bottleneck.
Karsten: How can Life Sciences companies get started if they are interested in generating more scientific insights in their engagements with customers?
Bernd: We usually start with a detailed briefing. In our project we have found that the set of initial questions very much affects the outcome of our projects. We work with our clients to narrow down general concerns into a set of clear, specific questions. PAI then analyses the data and provides the answers to those questions. We then iterate back and work on follow-up questions or further narrow down answers.
Karsten: Considering that different promoted brands go through changing phases of their life cycle, which of them can benefit most?
Bernd: Originally, we started out with late lifecycle brands since we thought that new data would help our clients to engage their HCPs given the relative lack of new studies for late lifecycle products.
We learned quickly that PAI has an application in every phase of the product lifecycle: We have helped clients in assessing indications for their market potential, studied market entry strategies, recommended product positions and provided content for late-lifecycle products.
Gradually - mostly driven by new questions from our clients - we have expanded the capabilities of PAI. For instance, we have added sentiment analysis to assess the mood of texts and we have expanded our data sources from just scientific studies to wider content from social media and even videos, which we automatically transcribe and analyse for content.
Karsten: How does such a sentiment analysis work and how do your clients review the outcome of such analysis?
Bernd: We have added sentiment analysis to answer questions on what the attitude towards certain topics is. We have a dictionary of words with their respective alignment: positive or negative. We use this to score texts. Our clients are often quite surprised by the sentiment results as it is hard to predict what the outcome is. Currently we are setting baselines with our clients as no one really has a sentiment score for discussions around their brands or products.
Illustration of sentiment analysis using PAI
We relate the identified topics via their respective distances to the words in the text. So, the topics are shown along the x-axis and the words (not all shown for clarity) along the y-axis. We can thus derive the distances of topics to each other. This is shown by the above dendrogram. Texts need to be pre-processed in order to reduce them to their stems. Words like “relapsing”, “relapse” are reduced to their stem “relaps” as shown on the y-axis. This is valuable information as clients are sometimes looking for special "island-topics" or the "general theme".
Karsten: Do you have another example where the use of PAI can be beneficial?
Bernd: Advisory Boards are a good example where PAI can be applied very nicely. In the frame of a market entry study, a client asked us to prepare its medical team for an Advisory Board. Based on the insights generated by PAI we briefed the client for his Advisory Board and provided a list of questions on key challenges in the indication. The Advisory Board was more valuable to the client as PAI provided the context upfront and the Advisory Board focused on the practical application of the scientific insights already collected.
Karsten: Where do you see PAI going in the future?
Bernd: We still see vast areas where we can add value with our existing technology. Data sources are an ongoing concern, not all data is publicly available. We are investing there. Also, clients still do not realize how much data they own, like CRM data or medical requests. Identifying and preparing data sources with the client is an area to be developed further.
Also, we are working on a fully automated content generator, mostly for digital channels.
Karsten: Thanks for sharing these interesting insights of the artificial intelligence world. In summary we can say that the use of PAI can be an important instrument to identify a significant amount of new content elements which otherwise would probably stay unnoticed.
What is your approach in identifying new relevant content in your therapeutic area? If you have further questions on this topic, please don’t hesitate to reach out to us at info@xeleratio.com or office@tegrum.eu.
Xeleratio Consulting Ltd.
We help Life Sciences executives improve sales performance with innovative best-in-class Business Excellence tools and methodologies . Expertise in Business Excellence has been gained with over 12 years of working in different global and regional roles in the Life Sciences industry.
Please feel free to reach out and request a free strategy session whenever convenient.
Comments