pjsample / human_5utr_modeling Goto Github PK

Great work! I think the model would be a valuable resource in Kipoi: model zoo for genomics (https://kipoi.org). Would you consider making a PR to https://github.com/kipoi/models with the model? I'm happy to help you with that.

Best
Ziga

modeling/training_MRL_CNN.ipynb

df.sort_values('total_reads', inplace=True, ascending=False)
df.reset_index(inplace=True, drop=True)
df = df.iloc[:280000]
e_test = df.iloc[:20000]
e_train = df.iloc[20000:]

The testing set has the highest 'total reads' values, which is not random splitting the data

If random splitting the data, the R^2 will decrease to ~0.88 (after removing UTRs with low 'total reads' using df = df.iloc[:280000]), or ~0.73 (not removing UTRs with low 'total reads' using df = df.iloc[:280000] )

Dear github author:
Recently I am interested in your work,and it is great. I have a problom,is that the .csv data I could find and download,but where is the .pkl data I could find? I am looking forward to your reply.

Best regards.

Hello,

To help users (me) get how to make effective use of Human_5utr_modeling, could you provide an input/output example set and the associated command line?
It would be really usefull!
Many thanks for any help you can provide me,

Véronique