Curation Support Tools for the Creation of a Lung Cancer Disease Map

[cannin]

Important Note

Look at Getting Started before asking questions.

Background

Similar to other professions, biologists need fast and reliable ways of summarizing massive amounts of research content found in literature. AI agents that produce detailed, factual, and unbiased research reports with citations are highly important. Tools like OpenAI Deep Research are steps in this direction along with open-source tools like the GPT-Researcher (https://github.com/assafelovic/gpt-researcher). One particular common type of report seeks to understand the biological networks involved in particular diseases or other processes. Communities like DiseaseMaps (https://disease-maps.io/projects/) seek to do this at the scale of diseases like Parkinson's disease which include both written material (https://pmc.ncbi.nlm.nih.gov/articles/PMC4153395/) and interactive diagrams (https://pdmap.uni.lu) that often become the source for mathematical models.

Goal

The goal of the project develop additional features or fine-tuning capabilities for gpt-researcher to support the automatic generation of reports for the curation of disease maps. The Related Links has a list of potential resources, including examples of existing reports and automatically generated reports as well as resources (e.g., EMMAA) that remain inaccessible to current state of the art tools like Deep Research.

Getting Started

1. Read the entirety of this Project Description.
2. Look at Getting Started, Goal and Related Links (especially Deep Research versus Human).
3. Compare GPT-Researcher and Deep Research in versus the human reports (Related Links).
4. Write a feasible proposal draft that outlines at some of the differences you see between automated and human reports and presents a plan to address this subset of differences.

Important: Avoid asking the generic questions such as how to start or if you can work on this project; just write a draft proposal if interested; draft feedback can be provided. Also, good proposals tend to have basic demonstrative code or point to specific code that will be enhanced/modified.

Difficulty Level: Medium

This is a more open-ended project; it will require a proposal that is focused enough to be achievable during the time frame.

Size and Length of Project

• medium: 175 hours
• 12 weeks

Skills

• Essential skills: Python
• Nice to have skills: Javascript

Related Links

• https://github.com/assafelovic/gpt-researcher
• Collections of automatically extracted interactions for diseases (including lung cancer): https://emmaa.indra.bio/
• Deep Research versus Human: Deep Research: https://drive.google.com/file/d/19z-PIz3JQLlOlv1wj-tFTgtK0px2-rqg/view?usp=sharing Human: https://pmc.ncbi.nlm.nih.gov/articles/PMC3898398/
• Example Reports:
• https://disease-maps.io/projects/
• Lung Cancer Report (try to replicate this): https://link.springer.com/article/10.1007/s00335-025-10110-6
• SIRT1/PARP1: https://pmc.ncbi.nlm.nih.gov/articles/PMC3898398/

Potential Mentors

Anbumathi Palanisamy
Augustin Luna

[atityam]

Hello @cannin @khanspers and entire team !!

I’m Aditya Maheshwari, an IT student with experience in Python and JavaScript, and I am highly interested in contributing to the "Curation Support Tools for the Creation of a Lung Cancer Disease Map" project for GSOC 2025.

I find this project particularly exciting because of its potential to enhance AI-driven research curation. The comparison between Deep Research, GPT-Researcher, and human reports highlights key challenges in automated literature analysis, such as maintaining factual accuracy, unbiased reporting, and structured citations. I would love to explore how fine-tuning or additional features in GPT-Researcher can bridge these gaps.

I have experience as a Front-End Developer Intern, where I worked on web-based projects, and I have also received an internship certificate for my contributions. My technical background includes working with Python, JavaScript, HTML, CSS, and Node.js, along with a keen interest in AI-driven research tools.

I have started analyzing the differences between human-written reports and AI-generated ones and plan to propose a targeted approach for improvement. Specifically, I aim to:

Identify where GPT-Researcher underperforms compared to human reports (e.g., citation accuracy, contextual reasoning).
Explore fine-tuning techniques to enhance disease map curation, possibly leveraging EMMAA and structured knowledge extraction.
Improve report structuring and reliability using AI-assisted summarization.
I would love to discuss potential directions for contribution and get feedback on my initial ideas. Please let me know how I can best align my work with the project's goals!

You can reach me at:
Email: adityamaheshwari5562@gmail.com
LinkedIn: https://www.linkedin.com/in/adityamaheshwarii/
Contact: +91-9893166465

Aditya Maheshwari-Internship Certificate.pdf

Looking forward to your guidance!

[ajaygaur01]

👋 Hello everyone,

I’m Ajay Gaur, a full-stack developer excited to contribute to this project! I have experience in Next.js, Prisma, PostgreSQL, AWS, and API integrations and am eager to apply my skills to improve AI-powered research tools.

I’m particularly interested in enhancing GPT-Researcher for automated disease research reports. While I am new to Python, I am actively learning and focusing on:

Improving report formatting and structure for better readability.
Integrating new data sources to enhance the accuracy of AI-generated reports.
Exploring AI fine-tuning to bridge the gap between human and automated reports.
Looking forward to collaborating and contributing to this exciting project! 🚀

Thanks, and happy to connect with everyone. 😊

GitHub: https://github.com/ajaygaur01

[lareinahu-2023]

Hi @cannin @khanspers ,

I'm Jiahui Hu, a Master's student in Computer Software Engineering . I'm very interested in contributing to this GPT-researcher enhancement project for disease map reporting.

My relevant background:

• Developed an NLP-powered search system with fuzzy matching to handle imprecise text inputs - directly applicable to improving how biological relationships are extracted from research papers
• Studied AI generative models in my coursework "Theory & Practice of AI Gen Model"
• Built systems with efficient caching layers that reduced database load by 40%, which could benefit processing large research datasets

After reviewing the example outputs in your shared links (Deep Research vs human-generated reports), I see opportunities to improve how automated systems handle specialized biological terminology and network relationships. The challenge reminds me of text similarity problems I've solved in my projects.

I'm curious about:

1. Are there specific biological knowledge databases you're considering integrating beyond EMMAA?
2. Have you explored fuzzy matching techniques to help the system recognize variations in how biological concepts are described?

I'd love to discuss how my NLP experience could contribute to this project. I'm excited about the potential impact better automated research tools could have on disease mapping.

Thanks,
Jiahui
Email: lareinahuwork@gmail.com

[Jiya873]

Hi @cannin,

I’m Jiya Gupta, a student with experience in Python and a keen interest in AI-driven research tools. I’m excited about the Curation Support Tools for the Creation of a Lung Cancer Disease Map project and would love to contribute.

Could you please suggest the best starting point for contribution? Should I begin by exploring GPT-Researcher, reviewing existing reports, or focusing on specific enhancements?

Looking forward to your guidance!

[cannin]

@Jiya873 see Getting Started

[aryan0931]

Hi, I’m Aryan Yadav, a third-year Mechanical Engineering student at IIT BHU, Varanasi. I have a strong background in AI, NLP, and Large Language Models (LLMs), and I’ve been actively contributing to open-source projects focused on AI-driven automation, data processing, and research optimization.

I’ve also participated in the LFX Mentorship Program, where I worked on optimizing LLM inference and integrating AI models in cloud-edge environments, collaborating with mentors across the US, Europe, and China. This experience gave me the opportunity to contribute to open-source tools and work in a collaborative environment focused on solving real-world problems. Additionally, I have gained valuable experience as a Data Science Intern at IIFL Home Loans, where I worked on predictive analytics, customer segmentation, and data-driven decision-making.

I’m eager to bring my skills and experience to the project and contribute to its success, while collaborating with the team to make a meaningful impact in the field of biomedical research.

Thanks,
Aryan Yadav
Email- aryan.yadav.mec22@itbhu.ac.in

[DebmalyaSen34]

Hello everyone!

I'm Debmalya Sen, a pre-final year B.Tech Biotechnology student at NIT Warangal. I'm very interested in contributing to this exciting project aimed at enhancing GPT-Researcher for biological disease map curation.

After carefully reviewing the project description and comparing automated reports (GPT-Researcher/Deep Research) with expert-written human reports, I've identified two key areas where improvements can significantly enhance performance:

• Integration of Specialized Biological Databases:
  Developing new retrievers within GPT-Researcher to access specialized biological databases such as EMMAA, KEGG, Reactome, and STRING. This integration will greatly enrich the biological accuracy and depth of generated reports.
• Fine-Tuning GPT Models on Biomedical Literature:
  Fine-tuning the underlying LLM specifically on biomedical datasets (e.g., PubMed abstracts/articles) to improve domain-specific understanding, precision of molecular interactions, and citation accuracy.

I have already begun exploring the GPT-Researcher codebase thoroughly to identify specific modules suitable for these enhancements. Given that fine-tuning will be a significant component of the project, I plan to prioritize this task in the initial stages.

Additionally, I have relevant prior experience: during my recent internship, I was responsible for creating biological interaction networks focused on cancer research and identifying significant genes involved. This work led us to draft a research paper currently under communication.

I'm excited about the opportunity to collaborate with mentors and community members on this innovative project.

Looking forward to contributing and learning together!

Thank you reading this long comment 😅

Best regards,
Debmalya Sen
mailto:ujjwalsen005@gmail.com
Github Profile

[M-Haseeb01]

Hi, I'm really excited about this project and would love to contribute! I have a solid foundation in Python and am eager to help enhance the tool to better support disease map reports.
Email:mhaseeb786786@outlook.com

[aparnasingh03]

Hey @cannin! I’ve read the Getting Started guide, and I just have a quick question. How can we get our proposal reviewed?

[cannin]

@aparnasingh03 if you want feedback on a proposal send me a draft at aluna@broadinstitute.org