作者回复：if families are changing too rapidly, it becomes hard to infer their most likely ancestral state, and therefore whether there have been gains or losses on a specific branch. But if you’d like to get a non-exact idea of what’s going on, you could try analyzing these families while setting lambda to a very low value (for instance, 0.0001)—do not search for lambda. This will tell you what the ancestral sizes were under a model where the families are evolving slowly, but should be an approximation of their true size.
#可以过滤掉这些基因家族, 再做分析
python3 /share/work/biosoft/CAFE/CAFE5/tutorial/clade_and_size_filter.py -i input.tab -o input.tab.filtered -s
###如果报错：Families with largest size differentials: https://www.omicsclass.com/question/5581
cat cafe5.log |grep "^OG" |awk -F":" '{print $1}' >family.remove.id.txt
grep -v -f family.remove.id.txt input.tab >input.tab.new
#再次运行
cafe5 --infile input.tab.new --tree input.tree --output_prefix cafe_result --cores 20 -k 3 -p --pvalue 0.05 &>cafe5.log1
