How to filter uninformative GO Terms

[Rohit-Satyam]

Hi!! Thanks for this easy to install and easy to use CLI tool for functional identification. I have few queries:

1. Can I use proteinfer for organism ranging from Bacteria to Eukaryotes? If so do we need to adjust any parameters?
2. Do you recommend running proteinfer with --num_ensemble_elements 5. I tested this parameter for few well annotated protein in my organism I saw that using 5 rather than default decreased the probability of the actual function. When do you recommend using ensemble parameter and when default works just fine?
3. Some GO Terms such as ones given below are not very informative and assigned high confidence score. For handful of genes it can be removed manually. However, for 100 and 1000s of genes manual removal is difficult. Is there a strategy to get rid of such terms?

   cellular_component
   biological_process
   cell part
   molecular_function
   intracellular part
   cellular process
   binding
   biological regulation
   cytoplasmic part
   organelle part

[Rohit-Satyam]

I wrote a short code in R to do that:

library(dplyr)
library(plyr)
library(stringr)

t <- read.csv("file.tsv",sep = "\t")
rm <- c("cellular_component","biological_process","localization","cell part","molecular_function","intracellular part","cellular process","binding","biological regulation","cytoplasmic part","organelle part","regulation of biological process","chromosomal part")
t <- t[!t$description %in% rm,]
t$description <- paste0(t$description," (",t$confidence,")")
t <- t[,-3]

temp <- t %>%
  group_by(sequence_name) %>%
  dplyr::summarize(label=str_c(predicted_label,collapse=","),desc = str_c(description, collapse = ", "))
write.table(temp,"filenew.tsv",sep = "\t")