Prostate Cancer Exploration Assistant

Inspiration

The inspiration for the Prostate Cancer Exploration Assistant stems from the need to simplify and enhance the exploration of resistance mechanisms in prostate cancer, specifically AR-V7. By integrating AI-driven insights and data retrieval, we can streamline research and improve understanding of how resistance to therapies develops. The use of Mistral LLM and Cortex AI ensures that researchers have quick access to the most relevant documents and insights, bridging the gap between data analysis and actionable knowledge.

What it does

The Prostate Cancer Exploration Assistant allows users to query a database of documents to explore mechanisms of resistance in prostate cancer, particularly related to AR-V7. It retrieves relevant documents via Cortex AI search, generating insights using the Mistral LLM. The app also logs user queries and generated insights for feedback and further analysis. Users can input their own context for personalized insight generation or rely on automatically retrieved documents for a data-driven approach to research.

How we built it

We utilized Streamlit for the front-end interface, providing a user-friendly experience for interacting with the app. The back-end leverages Snowflake Snowpark for querying the database and integrating Cortex AI for document retrieval, followed by Mistral LLM for generating insights based on the retrieved context. By combining document retrieval with AI-powered insights, the system enables efficient exploration of research topics and tracks user feedback for continuous improvement.

Challenges we ran into

Some challenges encountered during development included ensuring seamless fallback mechanisms between Cortex AI search and traditional SQL queries when no results were retrieved from AI search. Additionally, generating accurate insights that align with user expectations posed difficulties, especially when the context retrieved from documents was not ideal. Balancing system performance, real-time document retrieval, and AI-powered insights also required careful optimization.

Accomplishments that we're proud of

We are proud of successfully integrating Cortex AI for document search and Mistral LLM for generating high-quality insights, which not only enhances the usability of the app but also provides valuable information on prostate cancer resistance mechanisms. Additionally, the system logs insights and feedback, enabling iterative improvements and tracking user satisfaction. The app’s flexibility in allowing users to input custom context or rely on automated data retrieval is a key accomplishment.

What we learned

We learned a great deal about integrating multiple AI models and database technologies into a seamless, user-friendly application. One key takeaway was understanding the importance of designing fallback mechanisms and ensuring the quality of AI-generated content. Additionally, managing large datasets from a clinical or research-oriented database presented challenges related to performance and relevance of results. We also gained insight into the value of collecting user feedback for refining AI outputs over time.

What's next for Prostate Cancer Exploration Assistant

Moving forward, we plan to:

• Enhance the Cortex AI search capabilities to improve the relevance of retrieved documents.

• Improve the insight generation process by fine-tuning the Mistral LLM prompts and responses for better relevance and accuracy.

  • Explore integrating more datasets to broaden the scope of research and resistance mechanisms covered.

• Investigate the potential to integrate other AI models or scientific resources for a more comprehensive toolset.

• Optimize the system to handle larger datasets and provide real-time, actionable insights for researchers.

Built With

• cortex
• mistral
• python
• streamlit