Skip to main content
Fig. 4 | BMC Bioinformatics

Fig. 4

From: CysPresso: a classification model utilizing deep learning protein representations to predict recombinant expression of cysteine-dense peptides

Fig. 4

Features important for predicting expressibility in non-knottins are different from the features that are important for predicting expressibility in knottin peptides. This figure shows absolute SHAP values, which illustrate to what degree a feature affects prediction, calculated for different AlphaFold2 features. A The relative importance of amino acid position for prediction of expressibility in non-knottin and knottin peptides. In non-knottins, a region near the C-terminus (position 42 to 47) was identified as playing an important role in determining expressibility. In knottin peptides, the features at amino acid position 7–9 near the N-terminus were most important for expressibility. B The relative importance of the different AlphaFold2 representations for expressibility prediction of non-knottin and knottin peptides. For non-knottins, the structure representation module representation was most important for prediction of expressibility. On the other hand, in knottin peptides, the abstract single representation was most important for prediction of expressibility

Back to article page