Want to spend an entire decade surfing the web?
Try finding and reading every article on “artificial intelligence.”
A serious branch of computer science has rapidly turned into a buzzword used by bullish stock analysts, spammers, and the like.
Which is unfortunate, considering that we now have all of the ingredients in place to make artificial intelligence truly useful and practical.
We will leave it to others to discuss the basics of artificial intelligence. Good descriptions and discussions can be found here, here,here, and here.
The key feature of AI is its ability to learn based on past experience.
Complex algorithms can be trained to learn and provide results that are most useful to you. As a simple example, an online shopping web site will show ads for products based on your purchasing history.
In other words, AI is not a robot that performs repetitive tasks without variation. In a sense, AI is the opposite.
AI looks for patterns (based on predetermined rules, like shopping history) and proposes different “tasks” or solutions (like similar products) based on the inputs being analyzed by the algorithms.
So chess-playing AI systems become increasingly better at chess via data input (past chess matches) and by experience (playing more matches).
AI in Drug Discovery
Drug discovery is the perfect field for the application of AI. Indeed, a number of companies have already emerged which are focused on in silico methodologies and technologies for drug discoveries. Good overviews and examples of AI in drug discovery can be found here,here, and here.
The full potential of AI in drug discovery is still not understood, but it appears to be transformational. Some analysts suggest that AI can cut drug discovery time and costs by as much as 70%.
This is an astounding claim.
But if it is even remotely true, we may be on the tip of a complete transformation in how drug discovery is performed, in addition to vastly improved speed and possibly lower cost.
AI in Business Development?
Does AI have any application in business development and licensing?
Absolutely.
To answer this question, we look at three general phases of business development from an in-licensing perspective.
Search and Screen
Consider the different ways a company seeking in-licensing opportunities can go about finding opportunities:
- Unsolicited opportunities submitted via a company web site
- Attendance at partnering conferences
- Attendance at scientific conferences
- Attendance at clinically-oriented conferences
- Attendance at investor conferences
- Articles in journals, industry publications, newsletters, and social media
- Database searches
- Tips and referrals from friends and colleagues
Get the picture?
The fundamental issue here is that all of these disparate sources may have different companies and technologies listed in different formats, or with information behind paywalls or other closed environments.
For example, a company seeking a partner may post a detailed description of their out-licensing candidate in preparation for a partnering conference. However, the only way to view that content is to actually register for the conference and (with a little luck) search and find the information.
How else can a prospective partner find that information without attending that particular conference?
Company web site? Maybe, but how to find that web site in the first place?
We can go on and on, but the key point here is that company and drug candidate information is both everywhere and nowhere.
So where does AI come in?
Imagine having a tool which can search all of these data sources continuously…pulling information into an easy-to-interpret dashboard.
Now imagine the in-licensing team “training” the AI algorithms to automatically accept or reject opportunities based on key (and evolving) criteria.
Now imagine pulling other information that can be used to quickly advance or reject opportunities which do not fit the core criteria…such as trial information from clinicaltrials.gov, or patent information from uspto.gov…
Because of the paywall issue (coupled with different data formats), nearly of this is performed manually. But AI might be a great method to pull data from different public sources to aid in the search process.
Even processing inbound meeting invitations for a single conference can be a time-consuming process for a large, multinational company seeking licensing opportunities across multiple therapeutic areas.
AI would be a perfect system to have between partnering software and scouting teams to help process requests and focus only on those which meet key criteria.
Initial Evaluation and Prioritization
This is the area where AI can really provide value.
The Search aspects of AI are useless unless there is a way to quickly evaluate and prioritize all of the opportunities.
Again, this is a perfect scenario for AI.
Like grandmasters training a chess algorithm, scouts can also “teach” the AI system to evaluate and prioritize based on additional data, such as disease epidemiology, current therapies, pipeline data, and the like.
Imagine generating a detailed market assessment and revenue forecast with a few clicks.
Or how about pulling chemical structures of similar molecules and comparing drug-target binding data, based on published data from PubMed.
It may sound fanciful, but the data are already in place. There is nothing here that is not being done manually today.
But AI, coupled with solid data access, can pull these data, analyze them, and present them in readily digestible formats.
We will return to the “data access” issue later.
Diligence
Diligence simply takes the aforementioned model and expands on it for a select number of in-licensing opportunities. Algorithms could be taught to advance opportunities which meet or exceed minimum criteria, such as first in class, small molecule versus biologic, route of administration, and the like.
At some point, AI will turn the process over to the team, where experience, judgement, corporate, and “softer” factors will play a critical role in the in-licensing process.
The Data Access Issue
Many of the data sources needed are already publically available, such as text versions of publications in PubMed, patent text, epidemiology, web sites, social media, and so on.
But many sources are not.
And some may be publically available, but may be embedded in pictures, images, PDFs, or other file formats which are not easily machine readable. (Interestingly, voice and video recordings can be converted to text, and can then be searchable).
Companies using AI may also want to include their own private experimental data, which can be used to perform digital comparisons as part of an assessment.
As more and more companies realize this potential, they will likely make sure their non-confidential information is easy to find and easy to read by AI.
Better still, as AI formats emerge, we can envision companies taking all of their key non-confidential information and saving it in formats specifically for AI; like a vCARD for out-licensing assets.
Conclusion
AI may sound like a futuristic buzzword, but the reality is that it is already here.
In our industry, AI makes a great deal of sense for drug discovery, where managing large amounts of data and extracting insights through machine training and learning can lead to new innovations.
BD is no different.
Gathering and analyzing technical, commercial, legal, competitive, and other data categories currently employs dozens of analysts at pharma companies and consultancies around the world.
AI has the potential to replace much of this by continuously pulling data from public sources, analyzing it, and presenting it for human consumption and analysis.
As a result, scouts who spend time sourcing opportunities may spend more time feeding “private” data to the scouting algorithm (i.e., from conference partnering sites) and more time monitoring and evaluating the results presented by AI…teaching the algorithms to prioritize or deprioritize certain data aspects as more information is gathered.
Companies which provide AI services beyond drug discovery may be in an excellent position to lead our segment of the industry into this new age.
Attending BIO-Europe next month? Join us for a Pre-BIO-Europe gathering on Sunday, November 4, from 5 pm to 7 pm. We will meet at the Mikkeller Bar, which is world-famous for its hand-crafted brews. And it’s only 1 km away from the official Opening Reception at City Hall. Space is very limited, so grab a ticket (it’s free to attend) and have a few beverages and laughs with us before the conference.