Assisting the creation of SBGN diagrams using large language models (LLMs)

[hasanbalci]

Background

Systems Biology Graphical Notation (SBGN) is a standard language for representing biological pathways with SBGNML being the exchange and storage format for SBGN. SBGN diagrams are created using the developed SBGN editors (Newt, CellDesigner, SBGN-ED etc.) and they can be saved as SBGNML files. However, using these editors can sometimes be a difficult process, especially for those who try to use them for the first time, due to their complicated menus and toolbars. In addition, creating an SBGN map drawn on a paper or on a tablet during a meeting or lab discussion again to save it as SBGNML requires extra time and effort. To overcome these issues, we are developing a service and sample app (Image-to-SBGN Converter) to automatically convert hand-drawn SBGN maps to SBGNML by utilizing large language modes (LLMs). In addition to the conversion, this tool also allows user to see the resulting converted diagram and shows node-related information (see figure).

Goal

The goal of the project is to add some features to our app that will make the job of users easier before and after conversion and enable them to get more efficiency. These features include but are not limited to:

• Adding more conversion-related options to the Settings section
• Adding a quick edit feature if the result of the conversion has missing/wrong nodes/edges
• Providing functionality to open the converted map in an SBGN editor such as Newt for advanced editing
• Providing more detailed information about nodes and edges in Object View
• Adding layout options

Difficulty Level: Easy-Medium

Difficulty of the required features range from easy (making UI edits) to medium (using Cytoscape.js extensiones or querying data via REST API).

Size and Length of Project

• medium: 175 hours
• 12 weeks

Skills

Essential skills: JavaScript, HTML/CSS, Node.js
Nice to have skills: Cytoscape.js, LLM APIs (GPT, Gemini, Ollama etc.)

Public Repository

https://github.com/sciluna/image-to-sbgn

Potential Mentors

Hasan Balci
Augustin Luna

[NewtonPP]

Hello, My name is Prashidha Poudel. I am a sophomore computer engineering student at Mississippi State University. I have good experience with Javascript. I believe that I can highly contribute in this project,
Would we grateful if I get assigned to this project.

[Pritom2357]

Hi, I am Pritom Biswas.
I am a student in computer science and engineering in BUET, Bangladesh. I am well versed in Javascript, HTML, CSS, NodeJS and other frontend and backend utilities like databases. I am interested in this project and would like to contribute in here. i have went through your codebase and I think if I am assigned to this project, I will be able to help

[VatsalCodes44]

Hi, I’m Vatsal Mahajan from VIT Bhopal, and I have strong experience in JavaScript, HTML, CSS, and Node.js. I find this project exciting because it focuses on enhancing user experience and improving efficiency, which aligns with my expertise. I am particularly interested in implementing advanced editing features and working with tools like Cytoscape.js.

[sa778888]

Hi, I am Arshlaan Siddiquie
I am a student in computer science and engineering with bioinformatics . I am well versed in Python, LLM's , Javascript, HTML, CSS, NodeJS and other frontend and backend utilities like databases. I am interested in this project and would like to contribute in here. I would like to be assigned this task. Please guide me on this .

[virajranjan]

Hi, I am Vikrant Ranjan
I am an undergraduate student at NIT Durgapur with a strong foundation in biology, bioinformatics, and a keen interest in network biology. I'm drawn to the SBGN project due to its significant role in improving the accessibility and efficiency of SBGN-based pathway modeling by addressing key challenges in diagram creation. With solid skills in JavaScript in JavaScripts and LLMs, I'm eager to contribute effectively. I would greatly appreciate advice on how best to contribute and any specific area where you think my involvement could be most beneficial.