BARRNS & Genome

By integrating advanced computational tools and genomic data analysis, our app revolutionizes the clinical research landscape by accelerating the discovery and development of new treatments.

Comment

Inspiration

The inspiration behind the BARRNS & Genome App came from the need to streamline and accelerate the process of clinical research, particularly in the realm of genomics. As researchers ourselves, we've experienced firsthand the time-consuming nature of traditional genomic data analysis, including extensive literature reviews and manual data sorting. Observing the gap between genomic data generation and its validation and interpretation inspired us to create a solution that leverages the latest in computational power and AI technology. Our vision was to develop a tool that could not only process vast datasets efficiently but also integrate advanced analysis techniques that offer actionable insights, thereby expediting the path from research to clinical application.

What it does

The BARRNS & Genome App enhances genomic data analysis in clinical research by integrating advanced computational tools with the analytical capabilities of Large Language Models (LLMs). It efficiently processes gene quantification data from RNA-seq bioinformatics pipelines using DESeq2, manages large datasets, and generates critical visualizations like MA-plots, volcano plots, and gene counts. The app converts these visualizations into structured textual descriptions, allowing LLMs to provide detailed summaries, correlate findings with existing literature, and suggest actionable insights for clinical trials. This streamlined approach not only accelerates research but also ensures researchers focus on the most promising genetic candidates, thereby optimizing resources and enhancing research outcomes.

How we built it

The BARRNS & Genome App was developed through a collaborative effort between researchers, bioinformaticians, and software engineers. We utilized a robust stack of technologies, including:

  • Bioinformatics Pipeline: For data input, we incorporated outputs from bulk RNA-seq, using the popular bioinformatics tool DESeq2 for gene quantification analysis, chosen for its effectiveness in handling large datasets with limited computational resources.
  • Visualization Tools: We integrated tools to generate MA-plots, volcano plots, and gene counts directly within the app, and we are in the process of adding heat maps and PCA plots.
  • Large Language Model (LLM) Integration: We pioneered the integration of LLMs to process and interpret complex genomic data visualizations. This involved translating visual data into structured textual formats that an LLM can analyze, providing a textual summary and actionable insights based on the visualizations.

Challenges we ran into

Data Complexity: One of the main challenges was handling the complexity and sheer volume of genomic data effectively without overwhelming the computational resources.

Integration of LLMs: Developing a method to translate visual data into a format that can be processed by LLMs was challenging due to the multi-modal nature of the data.

Shiny App Development: Creating an intuitive user interface for the Shiny app posed initial challenges. Additionally, compiling all dependencies into a single repository was problematic; although using renv.lock facilitated environment replication, it didn't eliminate all issues. Hosting the app on shinyapps.io introduced a memory limitation, as the free tier provides only 1GB, insufficient for running complex analyses. Future solutions may involve upgrading to a higher tier membership to accommodate the increased memory demands.

What We Learned

Through this project, we gained insights into the capabilities and potential of integrating AI with genomic research. We learned about the intricacies of RNA-seq data analysis, the importance of effective visualization in conveying significant genetic expressions, and the transformative impact of LLMs in interpreting complex data. This project also highlighted the importance of interdisciplinary collaboration in tackling complex problems in the field of genomics and clinical research.

Team Culture

We enjoyed this project and our team had a great positive energy. We laughed a lot, had great collaboration, and our process developed over time as we each found our niche in different roles and responsibilities. As individuals with advanced backgrounds, we were able to offer our individual expertise, but more importantly, were able to use this hackathon as a way to explore other interests and passions.

What's next for BARRNS & Genome

  • Cloud Integration: We plan to scale our computational infrastructure by integrating with cloud services, which will allow us to handle data inputs at the BAM or FASTQ file stages, facilitating more comprehensive genomic analyses.

  • Expanded Visualization Capabilities: Adding more sophisticated visualizations like heat maps and dynamic PCA plots will provide deeper insights into the genomic data.

  • Enhanced LLM Capabilities: We aim to improve the LLM's algorithms for better accuracy and context-specific insights, potentially exploring the integration of newer AI models specialized in scientific data analysis.

  • Collaborations and Partnerships: To further refine and test our app, we plan to collaborate with pharmaceutical companies and laboratories, ensuring our tool meets their practical needs including accelerating clinical trials.

Built With

Share this project:

Updates