Utlizing the RA signature to predict response to TNFi