mgalardini
commented
8 months ago Can you try the following command: pyseer --kmers output_unitigs.pyseer --uncompressed --phenotypes train80_salt.pheno --wg enet --kmers output_unitigs.pyseer --save-model salt_unitigs.ridge --cpu 12 --alpha 0.01 > salt_selected_unitigs.txt, in which you are reading the kmers from scratch. I suspect that might cause the problem (which should throw a better error if that is the case). I would also perhaps generate k-mers/unitigs from the training set only, and not from the whole dataset, so that there is a stronger separation.
vijuvi
Hello!
I'm trying to use pyseer's elastic net feature to assess the effects of different variant groups on the phenotype, i.e. to find biologically plausible combinations of variants that predict the phenotype of the test set particularly well. I have randomly divided the set into training (80%) and test sets (20%). When I fit elastic net to the whole set, it works fine. When I try to train a model with the training set, it throws an error.
In other words, this command
pyseer --kmers output_unitigs.pyseer --uncompressed --phenotypes phenotypes/phenotypes.salt --wg enet --save-vars output/salt_unitigs --save-model salt_unitigs.ridge --cpu 12 --alpha 0.01 > salt_selected_unitigs.txtworks fine, but when I'm trying to fit a model to the training setpyseer --kmers output_unitigs.pyseer --uncompressed --phenotypes train80_salt.pheno --wg enet --load-vars output/salt_unitigs --save-model salt_unitigs.ridge --cpu 12 --alpha 0.01 > salt_selected_unitigs.txt, glmnet throws an error like this:Why does it say
Fitting elastic net to top 0 variantswhen with the whole set it saysFitting elastic net to top 324665 variants? An unitig compatibility issue?I'm using a conda environment with python 3.10.11 and these package versions: