Why autoscale (mean centre and divide by SD) before PLS-DA on integer data?

I'm currently working with a large dataset of integer data. I'm performing a variable stabilisation transformation on this data and then autoscaling before I begin working with PLS-DA and VIP. Can anyone explain to me why I've had to do this and how it's affecting my data? As I'm unclear myself other than the fact that it was recommended to me?