Predictor variable manipulation

Hello I was given a data set and asked to find the best statistical model to fit the data. I know how to sort through and choose predictor variables to include, but I am struggling with what manipulations I should make to the predictor variables to better fit the data. I ran into problems with the SSE and PRESS values being too far apart for me to suggest that it was an effective model. I have attached a picture of the residual plot which is currently using 2 of the 4 predictor variables. To reiterate my question is: based on the residual plot, what transformation should be made?