Closed cyber-ux closed 4 years ago
Hi Lucio,
Thanks for your interest.
extract
extracts features to text for training models or calling methylation. call_mods
calls methylation from fast5s or the text file generated by extract
.--write-path
can be changed to the path/file name whatever you want to.--model_path
must be set to the file path of the model. You can download the model from google drive. After uncompress it, the model can be used for call_mods
:
deepsignal call_mods --input_path fast5s.al/ --model_path model.CpG.R9.4_1D.human_hx1.bn17.sn360.v0.1.7+/bn_17.sn_360.epoch_9.ckpt --result_file fast5s.al.CpG.call_mods.tsv --reference_path GCF_000146045.2_R64_genomic.fna --corrected_group RawGenomeCorrected_001 --nproc 10 --is_gpu no
Best, Peng
So it's a standard / default template, if I don't download it I can't call the modifications ? This model is very large ? How much memory does it take up ? Then I have other questions : 1- I launched the feature extraction, I can't understand when the tool ends (the same thing for call_mod and frequencies), can you give me some examples to understand when the tool ends?. Also he tells me “write_process started”, what does it mean ? he tells me that : fast5_dir: fast5_single recursively: yes corrected_group: RawGenomeCorrected_000 basecall_subgroup: BaseCalled_template reference_path: GCF_000001405.39_GRCh38.p13_genomic.fna is_dna: yes normalize_method: mad methy_label: 1 kmer_len: 17 cent_signals_len: 360 motifs: CG mod_loc: 0 write_path: fast5s.al.CpG.signal_features.17bases.rawsignals_360.tsv nproc: 20 f5_batch_num: 100
14340000 fast5 files in total.. parse the motifs string.. read genome reference file.. write_process started.. Does that mean it's finished or is it still continuing ? 2- I saw in google drive and there are three models, based on what should I choose? Best
Hi Lucio,
The model takes 8G-11G GPU RAM.
1- write_process started..
means the extract
process just starts. deepsignal can't estimate when the process ends.
2- You can check README for the illustration of trained models. model.CpG.R9.4_1D.human_hx1.bn17.sn360.v0.1.7+.tar.gz
is our latest model for 5mCpG detection.
Best, Peng
Hi Peng, but does the model occupy 8G-11G once unzipped? How much does the compressed one take roughly? Best Regards
~500M disk space after unzipped.
So compressed is 500 Megabytes, decompressed 8-11 Gigabytes. Did I get it right ?
Not actually. The model takes ~500M space in disk after unzipped, and takes 8-11G RAM in GPU when calling methyaltion.
But before decompression how much does it take up ?
The model can't be loaded to GPU memory without unzipping it.
And to disk ?
Sorry, I cannot understand what you mean. The model tar.gz file is already in disk when you download it. It has to be unzipped and load to GPU for methylation calling.
What I ask is: when I download the tar.gz file of the model how much does it occupy on disk?
187M, the size of file can be checked in google drive.
Ok! Thank you so much for helping me! Best Regards
Hi Peng, I unzipped the model and it gives me a folder inside which it gives me 'bn_17.sn_360.epoch_9.ckpt.index' and 'bn_17.sn_360.epoch_9.ckpt.meta' but in the command instructions it is marked only '- model_path model.CpG.R9.4_1D.human_hx1.bn17.sn360.v0.1.7 + / bn_17.sn_360.epoch_9.ckpt '. My questions are : 1- Do I just enter 'model ..... ckpt' or do I enter 'model .... ckpt.index' or 'model .... ckpt.meta' ? I thought of inserting 'model ... ckpt.index', I think it's the right one, can you confirm? 2- What is the difference between ‘model...ckpt.index’ and ‘model...ckpt.meta’ ?
Hi Lucio,
Please refer to #11 .
Best, Peng
Hi, I would like some more information about this tool and therefore I would like to ask you: 1- what are the differences between extract features and call mods; 2- launching deepsignal as I can see when the tool ends and how long it takes 3- in extract features you can change the value of - -write-path or the one present is standard / default 4- in call modification you can change the value of - - model_path