You can calculate the percentage of NAs in each row and add that as a column to your data frame in R. Then you can use awk to remove rows that have more than 50% NAs.

In R:

count_na_func = function(x) {
sum(is.na(x))
}

df$na_percent = (apply(df, 1, count_na_func))/(ncol(df) - 1)
write.table(df, file = "datana.tsv", row.names=FALSE, sep="\t")

You can use some practice data to make sure that the na_percent column is computed correctly in R.

In bash:

awk '{if($12<=0.5){print}}' datana.tsv > newdata.tsv

Note that R is 1-indexed, while bash is 0-indexed. na_percent was in column 12 of my data frame, but it will probably be in a different column in yours, so you can substitute that column number for 12 in the awk code.