The full training data is missing

[karamveerverma37]

Hi, I would like to train a model for genome wide predictions and I found that the example dataset given is having a subset of training dataset (Chr19). Can you please share the full training dataset used for training.

[yupenghe]

Sure. See table S3 and S4 in the REPTILE paper
https://www.pnas.org/doi/full/10.1073/pnas.1618353114

[karamveerverma37]

Thanks for sharing the manuscript. As I can see this is raw data, can you share the preprocessed data used for training such as regions and their labels/state.

[yupenghe]

Will try but no promise since the data has been quite a few years. Also the training data will be on mm10 which probably won't help your case. Downloading raw data and reprocessing them on mm39 genome would be the best way.

[karamveerverma37]

Hi Thanks for the suggestion. My aim is to use trained models or train model in REPTILE to infer enhancers. I am not much aware of all types of data preprocessing. I have scripts to liftover from mm10 to mm39. Please share the data if available, I can use the preprocessed data used in the manuscript.

[yupenghe]

Ok. I think I got the training data. Will organize it a little bit before sharing. Do you read perl script (which is what I used to run training commands)?

[karamveerverma37]

Hi, thanks for the update. Yes I understand perl.

[karamveerverma37]

Also I would like to know the pretrained models provided in models directory. Are these models trained on the full genome data or only chr19. because when I try to use them to compute score I get the values only for chr19 and 0 for others chromosomes. If these are trained only on chr19. Can you share the pretrained models for full genome if available.
Thank you.

[yupenghe]

I try to use them to compute score I get the values only for chr19 and 0 for others chromosomes.

This is likely due to input files. Do you mind checking that the data of all chromosomes were used as input?

[yupenghe]

added the training and test data

https://github.com/yupenghe/REPTILE/tree/master/all_data

[karamveerverma37]

Thanks for sharing the training data. Does It also need bigwig files of the full genome, or it can take the data from epimark file.

[yupenghe]

It starts from epimark files.