1. I will first validate the data collections process and calculation logics to check if there is any bugs or technical issues that caused this problem. 2. If there are ok, it depends the fraction of th missing data: 1. if small fraction, drop that part of samples. 2., if too huge fraction, drop that whole columns