training a model for a new species

[mb47]

Hi,

thanks for providing QTG-Finder -- it sounds really promising and we would like to use in our current work. We work on barley and I was wondering if there are any instructions written down for how we would train a new model for this species?

From looking at the feature list, I think all the data required exist in some form for barley, but it would need to be pulled together appropriately, which could be a lot of work. Are there scripts for helping with getting this data together?

For example, it looks like we need a co-expression network, but what are the requirements for this? We have built barley expression networks with WGCNA before -- would these fit the bill? Or would the public BarleyNet resource be suitable (https://www.frontiersin.org/articles/10.3389/fpls.2020.00098/full)?

We also have an issue in that the various datasets are based on differing reference genome assemblies, so there would be a lot of liftover of features required potentially.

thanks
Micha

[fanlin-thermo]

HI Micha,
Thanks for the question.
In order to train a model in a new species like barley. You can use the scripts and example inputs to train a barley model.
https://github.com/carnegie/QTG_Finder/tree/master/QTG2_prediction/model_training

Some of the feature may not be available in barley, you can take them off and add your own features from public source like BarleyNet. The quality of may varies from databases to databases, I am not familiar with BarleyNet, so I could not judge if it is suitable.

Let me know if you have any other questions.

[Awais2426]

Hi,
Thank you very much for writing the QTG code to find orthologues in model species. I am new to Python and interested in running these codes to find the rice orthologue in Arabidopsis. I ran this code after downloading the input files and found the following error. Would you please help to sortout this issue.

(base) AwaisRiaz-01# python QTG_Finder_predict_orginal.py -gl AT_Seedsize_QTL_example.csv -sp "AT"
sklearn__version 1.5.2
Traceback (most recent call last):
File "/root/python/QTG_Finder_predict_orginal.py", line 72, in
df = pickle.load(pik_f) # Load the feature list with known causal genes used for training
^^^^^^^^^^^^^^^^^^
EOFError: Ran out of input
(base) AwaisRiaz-01#

Thank you

Awais Riaz