Create a program in Python that will run Principal Component Analysis in Google Colab on the data within train_cols that will be able to tell us which combination of variables are the best for a model's accuracy for test data. Assume that the variables within train_cols are your independent variables and "y" is your dependent variable.

train_cols = ["feature1", "feature2", "feature3", "feature4", "feature5", "feature6", "feature7", "feature8", "feature9", "feature10", "feature11", "feature12"]