“Harnessing Insights Across Clinical Trials with the Power of Conversation” — a presentation by Tobi Guennel, Ph.D., head of product development and innovation at QuartzBio
Watch the recording of the presentation:
Summary:
Discussions were lively at the “Digitization of Clinical Development & Clinical Trials” symposium at last week’s Bio-IT World conference.
Addressing an audience of biotech and pharma IT and data leaders, QuartzBio’s Tobi Guennel gave a provocative presentation on how deploying generative AI can amplify the ability of precision medicine approaches to accelerate drug development.
“It’s great to have a clinical trials-focused session at Bio-IT World, and being able to share success stories, especially about AI, in this landscape,” said Guennel. “We are firm believers that iteration and collaboration can only improve technology and create new paths for innovation.”
Guennel’s narrative centered around QuartzBio’s Biomarker Intelligence Platform, powered by AI, which allows precision medicine development teams to extract the most value from clinical trial samples and exploratory biomarker data.
“One benefit of natural language understanding capabilities of conversational AI is that a broad range of user personas, with diverse roles and functions, can interact with and extract insights from a unified, singular data ecosystem of sample and biomarker data.
Teams such as data science, translational research, and biosample or biomarker operations can now interrogate a unified data asset using natural language.”
Considerations for developing a large language model (LLM) ecosystem for precision medicine
To build a solid GAI framework to enable our sample and biomarker intelligence products, the QuartzBio team asked themselves:
- What LLMs are needed to support our domain and the tasks performed through the components of our products?
- How can we integrate these models into a scalable GAI workflow to support user stories and workflows without having to re-invent the wheel as we move from use case to use case?
Considerations for developing individual LLMs:
LLM Development
- Leverage existing Foundational Models: Start with existing powerful language models that are right-sized.
- Fine-Tuning: Customize these models for specific domains and tasks.
- Small LLMs: Create smaller, specialized models for targeted applications for improved cost/accuracy.
GAI Integration
- Prompt Engineering: Design tailored prompts to guide LLM behavior based on supported tasks.
- Leverage RAG: Combine LLMs with retrieval mechanisms for enhanced performance.
- User Agents: Implement user-specific agents to optimize model interactions and leverage live data.
Benefits
- Precision: Fine-tuned LLMs provide accurate and context-aware responses for specific tasks.
- Efficiency: Smaller models reduce computational overhead.
Supporting the entire precision medicine lifecycle
QuartzBio is integrating this GAI framework seamlessly into its Biomarker Intelligence platform to support a broad range of user stories and flows with a suite of SaaS products.
Sponsors are using the platform and products to build an interconnected data asset and, subsequently, draw insights via QuartzBio’s Biomarker Intelligence tools powered by conversational AI.
Sponsors leverage the platform as a force multiplier by creating internal efficiencies. Their teams are free to focus on insight generation rather than data wrangling.
Further, the QuartzBio platform amplifies knowledge by centralizing information and insights and making these easily consumable by a wide range of stakeholders. Ultimately, this increased access to intelligence enables sponsor teams to advance Precision Medicine objectives, such as accelerating patient selection strategies, identifying drug targets, and driving clinical trial efficiency.
Learn more: Watch a demo of QuartzBio’s AI-powered platform
Join the next webinar demonstration of our platform by signing up on our Webinars & Events page! We’ll invite you to our next webinar, and meanwhile you can watch one of our recent demos on demand.
