In vivo incorporation of non-natural amino acids can be used to improve protein stability. However, there is a trade off; improved stability of the protein may lead to loss in activity. One way to improve function is to employ machine-learning algorithms to identify the variants that improve activity. Our target protein Tetrahymena GCN5 (tGCN5), a member of the Histone Acetyltransferases (HAT) family, acetylates the lysine residues of histones, enabling transcriptional regulation. Experimental data have shown an increase in stability of the protein against protease but loss in activity with the incorporation of para-fluoro-phenylalainine (pFF) into tGCN5 in vivo. Using information from biochemical and structural data, we identified eleven residues to mutate. With the aid of computer guidance, we have designed a set of fluorinated variants and examine the activities.