Structure Prediction
Catana provides possibility to perform structure prediction based on the provided primary sequence. The prediction task itself is performed by the AlphaFold artifical intelligence service running on the Catana servers. From the user perspective, the prediction service is accessible via Structure - Predict menu and is thus realized mainly via the Structure predictor window shown below.
Using this window, the user can initiate a new prediction, check on status of running ones or retrieve (download or directly visualize) results of finished predictions.
IDs of submitted predictions are stored locally in the user's browser thanks to the HTML Web Storage API, so they are preserved when the browser is closed.
Moreover, the user can also provide an email address onto which an email notification will be sent after the prediction is finished.
Together with the PDB containing the predicted structure, so-called AlphaFold Score is retrieved. This score (formally called pLDDT score) determines the accuracy of prediction, while it holds that the higher the score is, the better is the outcome (for high accuracy predictions, score > 90 is to be expected). If you want to visualize the score on the structure itself, select "bfactor" color scheme and some linear color scale (for example, Orange-Red). Keep in mind that while the values are encoded as "B-factors", they in fact represent the AlphaFold Score and thus it holds that the higher, the better.
To better understand the evaluation of model confidence, we refer the reader to the publication of Jumper et al. 2021 and its Suppl. Methods 1.9.6.