search

l-magnificence / Mime

Machine learning-based integration model with elegant performance
Other
20 stars 7 forks source link

Lasso + GBM function error #19

Closed kerwin223 closed 1 week ago

kerwin223 commented 1 week ago

When running the Lasso + GBM model, the process gets interrupted: res <- ML.Dev.Prog.Sig( train_data = list_train_vali_Data$train_data, list_train_vali_Data = list_train_vali_Data, unicox.filter.for.candi = TRUE, unicox_p_cutoff = 0.05, candidate_genes = my_genelist, mode = 'all', nodesize = 5, seed = 5201314 ) 70e38c9556e3dd8f16b1de8c19111f9 2f5e78ab9bd21686d47f025d74fe214 e8563fd4e17bc04cc908d87daaabd49 35099f7dab8144f238029c89f4373f9

l-magnificence commented 1 week ago

I think this may be due to the fact that too few genes are selected for model construction. Can you add more genes for model construction and see if the error still exist.

kerwin223 commented 1 week ago

I think this may be due to the fact that too few genes are selected for model construction. Can you add more genes for model construction and see if the error still exist.

Appreciate！It works! Additional question: I tried building the model with 60 genes for the first time, but it failed. Is 60 the minimum number of genes required for the model?

l-magnificence commented 1 week ago

Because you set unicox_p_cutoff = 0.05, Mime will firstly filter genes associated with prognosis from your input gene list. If there are little genes meeting this criteria from 60 genes, it will be hard to construct models. By the way, you can also try set unicox.filter.for.candi=F to directly construct models from 60 genes.

kerwin223 commented 1 week ago

Because you set unicox_p_cutoff = 0.05, Mime will firstly filter genes associated with prognosis from your input gene list. If there are little genes meeting this criteria from 60 genes, it will be hard to construct models. By the way, you can also try set unicox.filter.for.candi=F to directly construct models from 60 genes.

Got it. Thanks for patiently explaining.