User Guide for DeepMBEnzy
If you do not have the SMILES representation of the compound, we recommend searching for it on PubChem (https://pubchem.ncbi.nlm.nih.gov/).Alternatively, you can click the “Draw molecule” button on the homepage to draw the molecule. Once drawn, right-click and select “Copy as SMILES,” then use Ctrl-C to copy the SMILES format.
For more information on drawing For detailed guidance on using the Molecular Editor (JSME), click the “JSME Help” button (https://jsme-editor.github.io/help.html).
DL Model Score: This is the SoftMax value generated by the model, indicating the probability that the candidate enzyme and reaction may occur.
Reaction Atom-Atom Mapping: This involves creating a one-to-one correspondence between all atoms in the reactant (left) and product (right). For more information, refer to Fooshee, David et al. "ReactionMap: An Efficient Atom-Mapping Algorithm for Chemical Reactions," Journal of Chemical Information and Modeling, 53(11), 2812-2819 (2013).
Mapping Attention to Enzyme: This feature extracts the model’s attention to the enzyme sequence and its three-dimensional structure. The original attention values and 3D structure can be downloaded for further analysis.
In summary, users should prioritize enzyme sequences with higher DL Model Score values for detailed analysis, considering reaction similarity and user expertise. Regions with higher attention values within the sequence are likely related to the enzyme's function and structure. These regions can be valuable for further analysis or experimental work, such as enzyme engineering.