Just to piggyback off this, one should be very careful combining techniques with PCA. One error I see a lot in biostatistics is people running statistical significance testing to select variables before running PCA. It's a super cool exercise to do that on datasets, because something WILL show up, and it's very often misleading.