Goosang-Yu
commented
6 months ago Oh... Thank you for uncovering this critical issue.
It seems that the problem arose due to the lack of code to convert the input sequence to .upper().
It appears that the issue arises when the input sequence contains lowercase letters, causing the GC count feature to be improperly counted, resulting in unexpected outputs.
# genet/predict/PrimeEditor.py/DeepPrimeGuideRNA
'GC_count_PBS' : [pbs.count('G') + pbs.count('C')],
'GC_count_RTT' : [rtt.count('G') + rtt.count('C')],
'GC_count_RT-PBS' : [self.rtpbs.count('G') + self.rtpbs.count('C')],Firstly, inputting all sequences in uppercase should yield the correct DeepPrime score. Could you please keep this issue open? I will work on a hot fix release as soon as possible to address the bug.
Thank you for finding and reporting this important bug!
P.S. I prefer this repo for discussing about GenET! Thank you.
francoiskroll
Is this the right repo to raise the issue? Or you prefer in DeepPrime repo? I'll put it here as this is where I got the example from documentation.
Input sequences in uppercase or lowercase give different outputs, which surely is not right...
Here is just taking the example from documentation:
From what I can tell, this issue does not seem to occur with
DeepPrime(...), i.e. whether wt_seq & mut_seq are lowercase or uppercase gives the same dataframe afterpegrna.predict(...).