PTMPredictionMetric

Reference

This is currently unpublished.

General description

Metric for predicting 18 different post-translational modifications (PTMs) using neural networks. The metric requires Rosetta to be build using extras=tensorflow (for compilation details see trRosettaProtocol). You can use this metric together with the GenericMonteCarloMover and RandomMutationMover to maximize/minimize the probability of a modification in your protein of interest.

Details and caveat

We use local sequence and structure features around the potentially modified site to predict its modification probability. We do not use global or homology features (e.g. whole sequence, ESM embeddings, MSAs, cellular localization), as the Metric is not only meant for prediction of PTMs but also for engineering PTMs of any protein, be it de novo or natural. This does mean that, for example, optimizing the probability of an N-linked glycosylation site will still not result in a glycosylated protein if the protein lacks a secretion tag or is expressed in an unsuitable system like E. coli.

Examples

Predicting N-linked glycosylation:

<ROSETTASCRIPTS>
    <FILTERS>
    </FILTERS>
    <RESIDUE_SELECTORS>
        <Index name="res" resnums="22A,38A,81A,165A,285A,63A,133A,144A,246A"/>
    </RESIDUE_SELECTORS>
    <SIMPLE_METRICS>
        <PTMPredictionMetric name="glycosylation_prediction" residue_selector="res" modification="NlinkedGlycosylation" />
    </SIMPLE_METRICS>
    <MOVERS>
        <RunSimpleMetrics name="run" metrics="glycosylation_prediction" override="true"/>
    </MOVERS>
    <PROTOCOLS>
        <Add mover_name="run"/>
    </PROTOCOLS>
</ROSETTASCRIPTS>

Autogenerated Tag Syntax Documentation:

A metric for estimating the probability of a given site to be modified, as predicted by neural network.

References and author information for the PTMPredictionMetric simple metric:

PTMPredictionMetric SimpleMetric's author(s): Moritz Ertelt, University of Leipzig moritz.ertelt@gmail.com

<PTMPredictionMetric name="(&string;)" custom_type="(&string;)"
        modification="(&string;)" residue_selector="(&string;)" />

custom_type: Allows multiple configured SimpleMetrics of a single type to be called in a single RunSimpleMetrics and SimpleMetricFeatures. The custom_type name will be added to the data tag in the scorefile or features database.
modification: (REQUIRED) The post-translational modification to predict. Available options are: Acetylation, ArgMethylation, Citrullination, Crotonylation, Deamidation, GammaCarboxyGlutamicAcid, Glutarylation, Glutathionylation, Hydroxylation, LysMethylation, Malonylation, NlinkedGlycosylation, OlinkedGlycosylation, Phosphorylation, SNitroysylation, Succinylation, Sumoylation, Ubiquitination
residue_selector: A residue selector specifying which residue or residues to predict on. The name of a previously declared residue selector or a logical expression of AND, NOT (!), OR, parentheses, and the names of previously declared residue selectors. Any capitalization of AND, NOT, and OR is accepted. An exclamation mark can be used instead of NOT. Boolean operators have their traditional priorities: NOT then AND then OR. For example, if selectors s1, s2, and s3 have been declared, you could write: 's1 or s2 and not s3' which would select a particular residue if that residue were selected by s1 or if it were selected by s2 but not by s3.

PTMPredictionMetric