heisenberg.ai

Home Page
Querying a Disease (Malaria) Brings You to The Generation Page
LLM Describes the Generated Compound
Entering a Question to The LLM
LLM Generated Answer

Inspiration...

Heisenberg.ai was inspired by the growing need to speed up drug discovery processes in the face of global health challenges. With advances in AI and machine learning, we saw an opportunity to leverage these technologies to create a platform that makes drug discovery more efficient, accessible, and innovative. The name "Heisenberg" pays pop-cultural homage to the scientific rigor of a pharmaceutical scientist.

What it does...

Heisenberg.ai is an artificial intelligence and machine learning platform that generates practical drug compounds for various diseases. Users can search for diseases, see a visualization of the generated medication in a 3D rendered view, and ask heisenberg.ai about the generated medication.

How we built it...

Heisenberg.ai was built using a combination of cutting-edge technologies:

Backend: FastAPI for high-performance handling of molecular data, SQLAlchemy for database management, and PostgreSQL for storing structured data. Frontend: Built with Next.js, React and Typescript to provide a seamless and responsive user experience, with 3Dmol.js for interactive molecular visualization. AI/ML: We integrated OpenAI’s GPT-4 for generating molecule descriptions and built custom machine learning models using SKLearn, leveraging the ChEMBL database to predict SMILES notations.

Challenges we ran into...

Throughout the development process, we learned the importance of blending AI with domain-specific knowledge, particularly in chemistry and drug discovery. Building a cohesive platform that could handle both complex scientific data and user-friendly interaction. We also gained deep insights into handling large databases, integrating machine learning models, and optimizing performance in a web-based environment.

Accomplishments that we're proud of...

We are proud of successfully implementing real-time rendering of a 3D molecule, which allows users to interact and explore the generated medication. Another major achievement was integrating OpenAI’s API which not only generates detailed drug descriptions. Our smooth disease search with autocomplete and autocorrect functionality is another feature that significantly enhances the user experience. As well as a cutting-edge UX/UI front-end design.

What we learned...

Throughout the development of Heisenberg.ai, we learned the importance of combining cutting-edge AI technology with domain-specific expertise in pharmacology and molecular biology. Integrating machine learning models into a practical application required a deep understanding of data structures, chemical notation, and molecular interactions. Additionally, we gained valuable insights into creating user-friendly interfaces for complex scientific tools, ensuring that the platform remained intuitive for users without sacrificing functionality. Collaboration between AI systems and scientific research tools presented challenges, but it also revealed the vast potential for AI to drive innovation in drug discovery.

What's next for heisenberg.ai?

We aim to scale Heisenberg.ai to generate thousands of potential new medications for a wide range of diseases. By expanding our machine learning models and integrating with larger, real-world chemical compound databases, we can enhance the accuracy and diversity of drug candidates. We also plan to implement features for predicting drug efficacy, safety profiles, and side effects, making the platform even more robust for research.