Fig. 5

The relationship between prediction quality, size of the neighborhood and used features. The dependence of the prediction quality on the size of the used c _k neighborhood (X-axis) and on the used features (individual lines). Shown are the following features types: amino acid type only (AA.AA), fingerprints with amino acid type and both amino acid type and RASA value on the central residue (AA.AA-RASA), RASA value only (RASA.RASA) and fingerprints with RASA value and both amino acid type and RASA value on the central residue (RASA.AA-RASA). Fingerprints length was 1023 bits and one most similar element was used