From a scientific question to a target shortlist: Meet the DeepLife Co-Scientist
Abstract
From a scientific question to a target shortlist: Meet the DeepLife Co-Scientist
Drug discovery increasingly relies on a growing ecosystem of tools: single-cell atlases, analytical pipelines, computational models, knowledge graphs and scientific literature. AI agents are beginning to connect these resources, but scientists still spend too much time navigating between tools, assembling evidence and testing hypotheses before they can confidently prioritize a therapeutic target.
The DeepLife Co-Scientist brings these capabilities together in one agentic platform, designed around the way drug discovery scientists work.
It connects DeepLife's single-cell atlases, curated interactome and knowledge graph, scientific literature, and TwinCell, our causal model, through a single conversational interface. Scientists can ask questions in plain language, while the Co-Scientist identifies relevant data, selects and runs the appropriate workflows, and delivers evidence-backed results that R&D teams can take forward to the lab.
From a question to a target shortlist
In the demo video, we walk through a target discovery workflow built around a real-world research question:
What is the best target to treat COVID-19 in the lung?
The Co-Scientist first identifies relevant datasets, selecting the Human Lung Cell Atlas for this example. It then prepares the appropriate disease contrasts and uses TwinCell to rank potential therapeutic targets.
But a ranked list is only the starting point. For each candidate, the Co-Scientist generates a target report showing the causal paths linking the target to genes affected by the disease, together with supporting scientific literature.
The workflow goes one step further: using the same disease contrast, it examines the targets of drugs already repurposed for COVID-19, including tocilizumab, dexamethasone and baricitinib. This provides an additional way to investigate the biological relevance of candidate targets and connect computational predictions with existing therapeutic knowledge.
Traceable workflows, reusable scientific skills
Trust and reproducibility are essential in drug discovery. That's why every Co-Scientist action is recorded in a lab notebook, capturing inputs, outputs and execution costs. Scientists can trace how results were generated, inspect the evidence behind them, and share the workflow with their teams.
Under the hood, each workflow is packaged as a reusable agentic skill. Our scientists can develop a workflow once and apply it across different diseases, cell types and partner projects. Target discovery is the first application, with indication expansion and additional workflows to follow.
Built for collaborative R&D
We already use the Co-Scientist in our research partnerships, working with partners' data and disease areas to accelerate the journey from a scientific question to an evidence-based target shortlist.
For R&D teams developing their own agentic systems, TwinCell can also be integrated directly through our API and the Co-Scientist's agentic skills. This allows in-house agents to access causal target evidence and combine it with existing internal tools and workflows.
Watch the demo to see how the DeepLife Co-Scientist turns a scientific question into a traceable, evidence-backed target discovery workflow.
Interested in exploring what it could do for your research? Get in touch to discuss a use case.




