Hello everyone,
I’m a bioinformatician looking to dive deeper into the world of machine learning and eventually deep learning. I have some basic theoretical knowledge of ML but no prior experience applying it in a biological context. To get started, I’d like to work on a learning-focused project using some RNA-seq data I have.
The data consists of bulk RNA sequencing from two cell lines originating from the same species but exhibiting a specific phenotypic difference. Ideally, I’d like to uncover what drives this phenotypic variation, but I’m also open to exploring any other project ideas using this dataset. My main goal is to learn and gain hands-on experience with ML techniques.
So far, I’ve processed the RNA-seq data, generating gene expression profiles for each cell line and a list of differentially expressed genes (DGE) comparing the two.
I’d greatly appreciate any suggestions on how to approach this project, including:
- Ideas for machine learning projects that fit this data.
- Recommended books or resources with hands-on code examples.
- Online courses or practical YouTube channels.
- GitHub repositories with well-documented pipelines and examples I can follow.
Thank you in advance for your help!
I feel like you're approaching this the wrong way around. Usually you find a problem or process that you think can be improved by ML, then find the datasets with appropriate controls and truth sets. Whereas here, you have a dataset of unknown quality and no problem or process to improve.
There are plenty of reviews that highlight areas of RNA-seq that can be improved with ML, especially in single-cell applications. So an alternative way of conducting this exercise could be to read these reviews, find a problem or process that interests you, and then search through data repositories for relevant high-quality data you could use.