Bo Wang co-authored a Nature paper demonstrating that flawed evaluation metrics caused linear models to appear to outperform nonlinear biological models