Great work!
In section 2.4.4 the dataset from Agarwal and Kelley, 2022 (https://genomebiology.biomedcentral.com/articles/10.1186/s13059-022-02811-x#availability-of-data-and-materials) is used but the mRNA stability model - Saluki from the same paper is not included in the benchmark. Saluki used UTRs and ORFs to make predictions so the comparison would ideally include a re-training/fine-tuning of the model on ORF only data but it would be interesting to see how the numbers compare to LLMs even without this (so, just using the existing Saluki model with ORF sequences only). I was wondering if there is a reason I am missing that lead to the choice of omitting this model from the benchmarking analysis.
Thanks!
Great work!
In section 2.4.4 the dataset from Agarwal and Kelley, 2022 (https://genomebiology.biomedcentral.com/articles/10.1186/s13059-022-02811-x#availability-of-data-and-materials) is used but the mRNA stability model - Saluki from the same paper is not included in the benchmark. Saluki used UTRs and ORFs to make predictions so the comparison would ideally include a re-training/fine-tuning of the model on ORF only data but it would be interesting to see how the numbers compare to LLMs even without this (so, just using the existing Saluki model with ORF sequences only). I was wondering if there is a reason I am missing that lead to the choice of omitting this model from the benchmarking analysis.
Thanks!