19F NMR chemical shift prediction with fluorine fingerprint descriptor

Abstract A novel strategy for 19 F chemical shift prediction is described. The approach is based on a new fluorine fingerprint descriptor and a distance-weighted k -nearest neighbors algorithm applied on a training set of known chemical shifts measured for different fluorine local chemical environments. It is simple, fast, accurate and interpretable, as it allows the user to compare the predicted chemical shift with the experimental chemical shifts of the neighbor structures, analyse the variability in their chemical shifts, and based on that have a knowledge-based assessment of the reliability of the prediction. Possible applications of this approach in combination with 19 F NMR-based screening in drug-discovery projects are discussed.