Staffing cuts and budget reductions are squeezing federal research agencies from both sides — yet your mission hasn’t gotten any smaller.
When critical reviews take 15–20 days, every lost day means slower funding decisions, higher risk exposure, and reduced program impact. Smaller teams simply can’t afford to waste time chasing data across siloed systems.
Waiting for resources to improve isn’t a strategy.
With fewer people to share the load, inefficiencies multiply — and so do the risks of missed impacts, unvetted partners, and misaligned funding.
Our new report, Doing More with Less: How Federal Research Agencies Are Maximizing Impact with Smarter Data Intelligence, reveals how agencies are:
Cutting review times by up to 90% — without adding headcount
Gaining real-time visibilityinto performance, partnerships, and risk
Reducing reliance on overburdened staff for manual data work
Securing data access in alignment with FedRAMP and DoD IL-4 requirements, pending 2026 certification
With Dimensions, your smaller team can work like a larger one — unifying publications, grants, patents, policy, collaborator data, and risk insights in one secure platform.
Get the report. Get the advantage.
Fill out the form to access your copy of Doing More with Less and see how other agencies are meeting higher expectations with fewer resources.
Doing More with Less: How Federal Research Agencies Are Maximizing Impact with Smarter Data Intelligence
This article distills key insights from the expert roundtable, “AI in Literature Reviews: Practical Strategies and Future Directions,” held in Boston on June 25 where a range of R&D professionals joined this roundtable, bringing perspectives from across the pharmaceutical and biotechnology landscape. Attendees included senior scientists, clinical development leads, and research informatics specialists, alongside experts working in translational medicine and pipeline strategy. Participants represented both global pharmaceutical companies and emerging biotechs, providing a balanced view of the challenges and opportunities shaping innovation in drug discovery and development.
Discussions covered real-world use cases, challenges in data quality and integration, and the evolving relationship between internal tooling and external AI platforms. The roundtable reflected both enthusiasm and realism about AI’s role in drug discovery – underscoring that real progress depends on high-quality data, strong governance, and tools designed with scientific nuance in mind. Trust, transparency, and reproducibility emerged as core pillars for building AI systems that can support meaningful research outcomes.
If you’re in an R&D role, whether in computational biology, informatics, or scientific strategy and looking to scale literature workflows in an AI-enabled world, keep reading for practical insights, cautionary flags, and ideas for future-proofing your approach.
Evolving Roles and Tooling Strategies
Participants emphasized the diversity of AI users across biopharma, distinguishing between computational biologists and bioinformaticians in terms of focus and tooling. While foundational tools like Copilot have proven useful, there’s a growing shift toward developing custom AI models for complex tasks such as protein structure prediction (e.g., ESM, AlphaFold).
AI adoption is unfolding both organically and strategically. Some teams are investing in internal infrastructure like company-wide chatbots and data-linking frameworks while navigating regulatory constraints around external tool usage. Many organizations have strict policies governing how proprietary data can be handled with AI, emphasizing the importance of controlled environments.
Several participants noted they work upstream from the literature, focusing more on protein design and sequencing. For these participants, AI is applied earlier in the R&D pipeline before findings appear in publications.
Stock image
Data: Abundance Meets Ambiguity
Attendees predominantly use public databases such as GeneBank and GISAID rather than relying on the literature. Yet issues persist: data quality, inconsistent ontologies, and a lack of structured metadata often require retraining public models with proprietary data. While vendors provide scholarly content through large knowledge models, trust in those outputs remains mixed. Raw, structured datasets (e.g., RNA-seq) are strongly preferred over derivative insights.
One participant described building an internal knowledge graph to examine drug–drug interactions, highlighting the challenges of aligning internal schemas and ontologies while ensuring data quality. Another shared how they incorporate open-source resources like Kimball and GBQBio into small molecule model development, with a focus on rigorous data annotation.
Several participants raised concerns about false positives in AI-driven search tools. One described experimenting with ChatGPT in research mode and the Rinsit platform, both of which struggled with precision. Another emphasized the need to surface metadata that identifies whether a publication is backed by accessible data, helping them avoid studies that offer visualizations without underlying datasets.
A recurring theme was the frustration with the academic community’s reluctance to share raw data, despite expectations to do so. As one participant noted:
“This is a competitive area—even in academia. No one wants to publish and then get scooped. It’s their bread and butter. The system is broken—that’s why we don’t have access to the raw data.”
When datasets aren’t linked in publications, some participants noted they often reach out to authors directly, though response rates are inconsistent. This highlights a broader unmet need: pharma companies are actively seeking high-quality datasets to supplement their models, especially beyond what’s available in subject-specific repositories.
Literature and the Need for Feedback Loops
Literature monitoring tools struggle with both accuracy and accessibility. Participants cited difficulties in filtering false positives and retrieving extractable raw data. While tools like ReadCube SLR allow for iterative, user-driven refinement, most platforms still lack persistent learning capabilities.
The absence of complete datasets in publications, often withheld due to competitive concerns, remains a significant obstacle. Attendees also raised concerns about AI-generated content contaminating future training data and discussed the legal complexities of using copyrighted materials.
As one participant noted:
“AI is generating so much content that it feeds back into itself. New AI systems are training on older AI outputs. You get less and less real content and more and more regurgitated material.”
Knowledge Graphs and the Future of Integration
Knowledge graphs were broadly recognized as essential for integrating and structuring disparate data sources. Although some attendees speculated that LLMs may eventually infer such relationships directly, the consensus was that knowledge graphs remain critical today. Companies like metaphacts are already applying ontologies to semantically index datasets, enabling more accurate, hallucination-free chatbot responses and deeper research analysis.
What’s Next: Trust, Metrics, and Metadata
Looking forward, participants advocated for AI outputs to include trust metrics, akin to statistical confidence scores, to assess reliability. Tools that index and surface supplementary materials were seen as essential for discovering usable data.
One participant explained:
“It would be valuable to have a confidence metric alongside rich metadata. If I’m exploring a hypothesis, I want to know not only what supports it, but also the types of data, for example, genetic, transcriptomic, proteomic, that are available. A tool that answers this kind of question and breaks down the response by data type would be incredibly useful. It should also indicate if supplementary data exists, what kind it is, and whether it’s been evaluated.”
Another emphasized:
“A trustworthiness metric would be highly useful. Papers often present conflicting or tentative claims, and it’s not always clear whether those are supported by data or based on assumptions. Ideally, we’d have tools that can assess not only the trustworthiness of a paper, but the reliability of individual statements.”
There was also recognition of the rich, though unvalidated, potential in preprints, particularly content from bioRxiv, which can offer valuable data not yet subjected to peer review.
Conclusion
The roundtable reflected both enthusiasm and realism about AI’s role in drug discovery. Real progress depends on high-quality data, strong governance, and tools designed with scientific nuance in mind. Trust, transparency, and reproducibility emerged as core pillars for building AI systems that can support meaningful research outcomes.
Digital Science: Enabling Trustworthy, Scalable AI in Drug Discovery
At Digital Science, our portfolio directly addresses the key challenges highlighted in this discussion.
ReadCube SLR offers auditable, feedback-driven literature review workflows that allow researchers to iteratively refine systematic searches.
Dimensions & metaphacts offers the Dimensions Knowledge Graph, a comprehensive, interlinked knowledge graph connecting internal data with public datasets (spanning publications, grants, clinical trials, etc.) and ontologies—ideal for powering structured, trustworthy AI models that support projects across the pharma value chain.
Altmetric identifies early signals of research attention and emerging trends, which can enhance model relevance and guide research prioritization.
For organizations pursuing centralized AI strategies, our products offer interoperable APIs and metadata-rich environments that integrate seamlessly with custom internal frameworks or LLM-driven systems. By embedding transparency, reproducibility, and structured insight into every tool, Digital Science helps computational biology teams build AI solutions they can trust.
