Agent-Assisted Coding: Best Practices for Bioinformatics

Overview

AI coding agents are tools that do more than suggest a line of code: they plan, edit files, run commands and iterate on the results. They are entering bioinformatics workflows quickly. They can speed up routine work such as writing parsers, debugging pipelines and adding tests, but they can also produce code that runs cleanly and is quietly wrong: a shifted genomic coordinate, a silently dropped record, a wrong constant.

This one-day, hands-on course teaches how to use coding agents in bioinformatics work effectively and responsibly. You will learn how agents work and how to frame tasks and give them the right context. You will then practice on realistic exercises built on synthetic data: parsing a file format, debugging a broken pipeline, and reviewing agent-written code that contains a hidden error. The central skill of the course is supervision, meaning verifying and reviewing what an agent produces rather than only prompting it. The course also covers data privacy, reproducibility and scientific integrity, and ends with a capstone in which you refactor, test and document a messy analysis script with an agent.

Audience

This course is designed for PhD students, postdoctoral and other researchers in the life sciences from both academia and industry who write code to analyse biological data and want to learn how to use AI coding agents in their work productively and safely. It is aimed at participants who are new to agentic coding tools.

Learning outcomes

At the end of the course, the participants should be able to:

• Explain what an AI coding agent is and how it works (model, harness, state, tools and the agentic loop)
• Frame tasks and provide context so that an agent produces useful code for common bioinformatics problems
• Use an agent for typical workflows such as parsing a bioinformatics file format and debugging a pipeline
• Review and verify agent-generated code, and detect silent errors such as coordinate-system off-by-one mistakes, dropped data or wrong constants
• Identify what must never be shared with an agent and apply good data-protection practice
• Document AI-assisted work so that it stays reproducible and scientifically sound

Prerequisites

Knowledge / competencies

This course is designed for users who already write some basic code for data analysis. There is no requirement for prior experience with AI coding tools. You should be comfortable with:

• Basic programming in Python (the exercises are in Python; R users who can read Python are welcome)
• Working on the command line
• Basic git usage Basic familiarity with common bioinformatics file formats (VCF, SAM, BED) is helpful but not required.

Technical

You will need access to a computer (Windows, macOS or Linux) on which you can install software and use a terminal. Python 3.10 or later, git, and the Python packages pandas and pytest are needed. You will also need a free personal account for the AI coding tool used during the course. Setup instructions and a short check script will be sent to registered participants before the course. Please complete the setup beforehand, so that the course time is spent on the exercises. All exercises use synthetic data. Please do not bring patient data or any other confidential data to the course.

Schedule – CE(S)T time zone

09:00 – 09:40 : Welcome and introduction: what AI coding agents are and how they work

09:40 – 10:20 : Context and task framing (hands-on)

10:20 – 10:40 : Break

10:40 – 11:25 : Workflow 1: parsing a bioinformatics file format (demo and exercise)

11:25 – 12:10 : Workflow 2: debugging a broken pipeline (demo and exercise)

12:10 – 13:10 : Lunch

13:10 – 13:55 : Verification and review of agent output (demo and exercise)

13:55 – 14:15 : Data privacy and security

14:15 – 14:35 : Reproducibility and scientific integrity

14:35 – 14:55 : Break

14:55 – 15:20 : Tooling landscape, limitations and failure modes

15:20 – 16:30 : Capstone: refactor, test and document an analysis script with an agent

16:30 – 17:00 : Wrap-up, open questions and course evaluation

Application

The registration fees for academics are 100CHF and 500 CHF for for-profit companies.

While participants are registered on a first come, first served basis, exceptions may be made to ensure diversity and equity, which may increase the time before your registration is confirmed.

Applications will close on 10/02/2027 or as soon as the places will be filled up. Cancellation after 10/02/2027 will not be reimbursed. Please note that participation in SIB courses is subject to our general conditions.

You will be informed by email of your registration confirmation. Upon reception of the confirmation email, participants will be asked to confirm attendance by paying the fees within 5 working days.

Venue and Time

The course will take place at the University of Lausanne (https://planete.unil.ch/plan/) (Metro M1 line, Sorge station).

Additional information

Coordination: Diana Marek, SIB Training group.

A Certificate of Attendance will be sent provided you were present at the course, whereas a Certificate of Achievement recommending 0.25 ECTS will be sent provided you passed the exam.

You are welcome to register to the SIB courses mailing list to be informed of all future courses and workshops, as well as all important deadlines using the form here.

SIB abides by the ELIXIR Code of Conduct. Participants of SIB courses are also required to abide by the same code.

For more information, please contact training@sib.swiss.