2024 was a BIG year for generative artificial intelligence (GenAI). Models, both public and commercial, made huge advancements against a broad set of benchmarks, with consumers seeing everyday exposure and accessibility skyrocket as well as associated costs plummet. Despite some ongoing challenges regarding ethical and security issues, the uptake of GenAI in numerous domains, from marketing and operations to healthcare and life sciences, was clear for all to see. Many efforts focussed on the application of GenAI in enabling retrieval augmented generation (RAG) backed question-answer systems, allowing users to ask questions of their data using natural language in an interactive fashion – pretty cool, eh? But what’s next?
We have already seen some game-changing announcements in the space of GenAI, with the launch of DeepSeek already being referred to as a Sputnik moment for AI. DeepSeek can only be a positive thing for consumers of commoditized GenAI, showing how more performant models can be built at a fraction of the cost using ‘modest’ hardware – emphasizing the trend of ever-decreasing costs. It also reminds us of the importance of ethical and transparency considerations – things that could well end up being the differentiator for the model, as well as service providers alike.
Everyone has experimented with use-case 101, chatbots, and attention moves to expediting more laborious human workflows. We have seen the rollout of numerous efforts from OpenAI that provide more than just ‘chat’ functionality. For example, o1 released in 2024, was capable of advanced reasoning, meaning it could think through complex problems and generate a chain of thought before providing an answer.
Fast forward to 2025, The Year of The Agents. From OpenAI, we have already seen the Operator, an agent for completing web-based tasks, and now Deep Research, an agentic capability that conducts multi-step research on the internet for complex tasks. The Year of Agents can be interpreted as the year people start extracting know-how from human subject matter experts (SMEs) and codifying this into their software. But where could this agentic AI bring value to the life sciences? And, importantly, what considerations should be made when looking at agentic AI?
When it comes to talking about agentic AI, it is often good to start with what we mean by agency, defined as the below:
We can extend the above to define AI agents, utilizing AI in a process or task where it is given some level of agency, i.e., the freedom to make decisions on behalf of a user who has delegated such responsibility. Agents are designed in such a fashion that they can complete specific tasks that may well take multiple steps to complete manually, such as target prioritization. A target prioritization agent, for example, will take a therapeutic area as input (e.g., Type 2 Diabetes) and provide a list of prioritized potential therapeutic targets (e.g., GLP1R ).
It makes sense to give agents the right tools or functions to complete the job.
For example, the target prioritization agent, which takes a disease area as input, will need to access:
We can think of the above as tools available to the agent, and these may well be external APIs or software; functions may also be needed to specify how data is to be manipulated or shared between tools or presented back to the end user.
With tools and functions defined, there are a few ways in which they can be made available to agents.
On one end of the spectrum, one may send the bag of jumbled tools and functions to a large language model (LLM), giving it the power to define which tools should be used and in which order. In this scenario, the LLM is not only the reasoning engine but also the planning module; the LLM has the autonomy to break down high-level goals into actionable steps for execution.
In this setup, you are ultimately giving the LLM the autonomy to plan and execute the execution of tasks.
On the other end of the spectrum, one may pass a prompt to an agent with a set of tools and functions along with a defined linear flow that the agent should follow to complete the task, e.g.,
Here, the LLM is not having to define the actionable steps, as they are already provided and defined, so is not the planning module, but it still brings reasoning to the process.
In addition to having truly autonomous vs guided agents, agents may be combined with other agents to complete even more complex tasks. For example, let’s say you also had an agent that specialized in identifying areas of unmet need, you may use that agent in conjunction with your target prioritization agent to look for prioritized targets for areas of unmet need, in what we call a multi-agent set up; again these may be shared as tools for an LLM to string together as it deems fit, or in a predefined linear fashion.*
Depending on the use case, components can be plumbed together in various flows with varying complexity. In addition to the tools and functions mentioned above, agents often require a few other components, including short-term memory and access to high-level descriptions of the tools they have at their disposal (e.g., “Use this service to review whether your protein target can be modified by small-molecule compounds”).
* Other notable setups include a mixture of expert agents (bringing together agents that are experts in differing fields to complete cross-domain tasks) and even swarms of agents (agents that can replicate or dynamically transform), along with ambient agents (who listen to an event stream and act on it accordingly, potentially acting on multiple events at a time) – but they are probably for another blog!
When it comes to the life sciences, particularly pharma drug development, there exists a whole raft of time-consuming tasks that are primed to be ‘agentified’, from R&D target identification, lead identification, and pre-clinical toxicology to clinical trial, approval, and post-marketing surveillance/line extension use-cases.
Agents present us with an excellent opportunity to expedite tasks that currently take the human (read scientist) time, with the potential to slash how long it takes, as well as the costs associated with getting drugs to market.
Having said that, the life sciences are a domain where evidence-based decision-making is paramount; wrong decisions can have costly and, in some cases, dire consequences. The idea of building a pipeline off the back of hypotheses generated from LLM-powered agents alone is something that should be, rightfully, approached with caution.
For agents to truly be beneficial, there are a few considerations that should be prioritized…
Start with the problem
Before developing any solution, be sure to truly understand the problem you are trying to solve. Refrain from trying to identify problems for which to apply your solution.
Transparency is paramount
For outcomes of agents to be trusted, traceability, and therefore transparency are key. It is not enough to say we are focussing on target X as part of our pipeline because agent Y told us to. It is vital to be able to understand WHY decisions were made – what tools were used, what data was reviewed, and what was the reasoning that led to the output from the agent?
Accessibility is key
For agents to consider the relevant search space when addressing a task, it is essential that they can access all necessary resources. This comes in the form of documented APIs presented to the agents as tools. Not only should resources be accessible with clear documentation, but there should also be some level of tool ‘description’ that can be interpreted, for example, “Need to know what the mechanism of action of a drug is? Then go to this service?”. One should also consider permissions of the data that is accessed by the agent.
Data rules
Although accessibility to resources is critical, outputs are only as good as the quality of the data that they have access to. This comes from having gold standard data for the domains in which you play.
Adding scientific meaning
LLMs are great at understanding language but can struggle to understand the synonymity and ambiguity that is ever-present in science; science is more complex than language.
For example, knowing that a mention of TA123 is the internal code for the gene GLP1R, is crucial when it comes to harmonizing data and amplifying pertinent signals.
Retrieving data from a single source becomes problematic without understanding the varied ways in which an entity can be referred to; for example, most open-source databases don’t have great semantic search and instead rely on basic synonyms. An agent will search with perhaps one synonym, e.g., Type 2 Diabetes, it will then assess results and, if it’s not satisfied (because for example, the results don’t have the right context), it may try other synonyms such as T2D and iterate around again – a slow and costly process. If, instead, the search tool was able to better deal with synonyms and higher-level concepts upfront, the search can be made a whole lot more efficient.
Furthermore, different resources can represent their data in multiple ways, both syntactically (e.g., unstructured text in search tools vs structured data in graph databases) as well as semantically (a search system may use MeSH to capture diseases, whereas your graph index may use the DOID) – having the ability to describe equivalency via mappings (e.g., MeSH:D003924 and DOID:9352 are both identifiers for the disease Type 2 Diabetes Mellitus).
By providing an LLM that understands language with scientific knowledge captured in ontologies, human explainability of outcomes can be provided as part of the transparency step.
Leave it to the LLM?… or use your SMEs?
Be careful how much autonomy you give your agent. Certain tasks will require little and may be better suited to a more linear process defined upfront. As opposed to trusting an LLM to decide the next step to take in a process, give it a clear plan, or even reach out to an SME for input and review before moving on to the next step in the execution plan.
There are scenarios where autonomy may be better suited, such as identifying novelty, but again, be sure to use your SMEs for evaluation.
Start with the smallest blocks
By focusing on having a clear set of tools or functions available, one can build a multitude of different agents atop. Think modular and document these as best you can. It is important to think bottom-up when developing agents.
Guard rails
Ensure that strict guard rails are in place when building out agents, and make it clear what the agent is not to do as much as it is to do!
Life Science problems do not require purely technical solutions
A complex technical setup does not equal fit for purpose. It is vital that the subject matter expert remains a key part of the development of any solution, in terms of design and critically, in their evaluation.
Agents often have access to multiple tools. These tools may well contain data that is captured in multiple technical or syntactic formats, such as an SQL database, a knowledge graph, or a document index accessible via an API. While LLMs are fantastic at translating language, for example, natural language to query language Y, such as SQL, cypher, or a RESTful API call (given the relevant documentation), they can struggle to convert natural language to the relevant identifiers used to describe the things of relevance within these systems.
Furthermore, in scenarios where different systems use different standards to represent the same things, it is vital to be able to harmonize the output from one system with the output of another, that is, to understand equivalence.
Ontologies can help alleviate all the issues highlighted above. Enriched versions of them can be used to convert natural language to identifiers, converting unstructured text to machine-readable identifiers when building databases. The same approach can be used to convert the string Glucagon-like peptide-1 to the ID GLP-1 for querying over a knowledge graph. Furthermore, ontologies can be used to capture equivalence where multiple standards exist to define the same domain, such as diseases in MeSH, DOID, ORDO to name a few.
The synergy between quality data, ontologies, and GenAI in the context of agentic AI provides a powerful toolkit that can be used in combination to provide time-saving solutions to the drug discovery process.
Agentic AI is here, and the potential value it will bring in supporting pharma R&D and the entire drug (read product) development pipeline is exciting for all to see. However, that should also be tempered with considerations discussed in this article. Like other technical approaches, agentic AI is still heavily dependent on the quality of the data it is presented with, which in turn is reliant on agents being able to locate, access, harmonize, and understand quality data – something that ontologies and the SciBite semantic product set can support.
As a part of Elsevier, SciBite are committed to leading this transformation by providing quality data, ontology-based data enrichment technologies, and expertise that can empower agentic AI applications throughout the life sciences.
Together, we can unlock a future where data-driven decisions are not just possible but are the norm in advancing drug discovery.
Leading SciBite’s data science and professional services team, Joe is dedicated to helping customers unlock the full potential of their data using SciBite’s semantic stack. Spearheading R&D initiatives within the team and pushing the boundaries of the possible. Joe’s expertise is rooted in a PhD from Newcastle University, focussing on novel computational approaches to drug repositioning; building atop semantic data integration, knowledge graph & data mining.
Since joining SciBite in 2017, Joe has been enthused by the rapid advancements in technology, particularly within AI. Recognizing the immense potential of AI, Joe combines this cutting-edge technology with SciBite’s core technologies to craft tailored, bespoke solutions that cater to diverse customer needs.
Other articles by Joe