Nonstandard values in DMS csv files

[alex-hh]

Hi, had a go at running the fitness benchmark and just wanted to flag that there are a few non-standard values in some of the datasets:

• some of the DMS files provide DNA sequences, some RNA sequences - is there a reason for this?
• all of the sequences in A0A2Z5U3Z0_9INFA_Doud_2016 end in 'EXI' - does this have a meaning?
• Soo_2021_ribozyme contains rows with NaN DMS_score values

Thanks again for the benchmarks!

[rohitarorayyc]

Thank you for your comments!

1. Only RNAs of the class mRNA-coding are in DNA sequence format, while the remainder of the RNAs are sequence format (https://github.com/MarksLab-DasLab/RNAGym/tree/main/fitness/reference_sheet_final.csv). It would be helpful to refer to the reference sheet for this info. We'll either make this clearer in the documentation or switch over all files to constant formatting.

2. Thanks for pointing this out, this is an error and we'll fix this! Will update when this is fixed here.

3. If you re-download the data, the Soo files should be fixed (NaN rows removed), see Missing Labels in Soo_2021_ribozyme.csv File #6 .