About Virus Watch

About Virus Watch

We as Innophore are proud to have teamed up with Amazon Web Services (AWS). The global DDI (Diagnostic Development Initiative) program, which supports SARS-CoV-2 projects worldwide, will provide us with computing power through AWS sevices to develop and handle predictions of upcoming SARS-CoV-2 variants based on structure and model information with our CatalophoreTM point-cloud technology. This is intended to support diagnostics and drug development by predicting potentially dangerous or possible new infectious variants, thereby providing some confidence in dealing with new variants of the pathogen.


What can I submit?

You can submit any sequence containing the RBD part of the spike protein. While checking the quality of your sequence, the RBD part will be detected and processed automatically. In case no RBD sequence is detected, you will be notified. In order to process you sequence, the sequence identity needs to be above 30%.

How long does the calculation take?

Since there are several steps involved, the calculation takes between 15 to 20 minutes. The process panel tracks the current status of your calculation. Please do not refresh the page while your calculation is performed.

My sequence did not pass the quality check, what should I do?

In order to make the sequence processable, it has to pass a quality check. Your sequence needs to have at least 80% of the original length compared to the wild-type RBD and 30% of sequence identity to the wild-type RBD sequence. Furthermore, your sequence can only contain letters coding for amino acids.

About Sequence Analysis


Your sequence will be submitted to a thorough check. That involves scanning for the RBD part and truncating your input to said part. Scanning for mutations by comparing the truncated sequence to it’s wild-type counterpart is performed to ensure the production of reliable results. Your sequence must be at least 30% identical to the wild type and must not contain unusual characters or non-standard amino acids. This is important for the model prediction. A summary of the quality check will be provided containing the RBD part of your submitted sequence and a list of the mutations present. After your sequence is checked and deemed fit for further processing, it will be used for the model building.

About 3D protein-structure prediction


A 3D structural model will be created of your RBD sequence in complex with its receptor, the human angiotensin-converting enzyme 2 (hACE2). This is achieved through homology modeling, where your RBD sequence is aligned to an experimentally solved crystal structure of the wild-type RBD in complex with hACE2. In the next step, these models are used as templates to calculate point clouds (Halos) from the binding interface of your predicted RBD-hACE2 complex employing our CatalophoreTM technology.

About CatalophoreTM Halo generation


Halos are our tool to calculate and depict the properties found on a surface of a protein. These properties are stored in point clouds that go beyond the surface. Each point of such a point cloud represents a property value at a given place in 3D space. To generate the input for the affinity prediction, Halos for both proteins involved in binding (the RBD and ACE-2 human receptor) need to be generated. Even though the calculation of a broad variety of properties is possible, for this experiment hydrophobicity, hydrogen donors, electric field and aromatic field are taken into account.

About Deep-learning based affinity prediction (ΔG)


Halos generated in the previous step are then transformed into a voxelized/volumetric cube with said properties. This information is fed into a three-dimensional convolutional neural network (CNN) architecture. This network consists of two steps, the first being applied individually on both input Halos, before they are joined by a "zipper" convolution and then processed jointly by another CNN to produce a prediction. The calculated binding affinity is represented by the ΔG value.

Further Reading

More background information on the tool can be found in our recent publications in Scientific Reports:

Optimizing variant-specific therapeutic SARS-CoV-2 decoys using deep-learning-guided molecular dynamic simulations.
K. Köchl, T. Schopper, V. Durmaz, L. Parigger, A. Singh, A. Krassnigg, M. Cespugli, W. Wu, X. Yang, Y. Zhang, W. Wen-Shang Wang, C. Selluski, T. Zhao, X. Zhang, C. Bai, L. Lin, Y. Hu, Z. Xie, Z. Zhang, J. Yan, K. Zatloukal, K. Gruber, G. Steinkellner & C. C. Gruber Sci Rep.

Structural bioinformatics analysis of SARS-CoV-2 variants reveals higher hACE2 receptor binding affinity for Omicron B.1.1.529 spike RBD compared to wild type reference.
V. Durmaz, K. Köchl, A. Krassnigg, L. Parigger, M. Hetmann, A. Singh, D. Nutz, A. Korsunsky, U. Kahler, C. König, L. Chang, M. Krebs, R. Bassetto, T. Pavkov-Keller, V. Resch, K. Gruber, G. Steinkellner & C. C. Gruber Sci Rep 12, 14534 (2022).

Your session-ID:
Am Eisernen Tor 3 | 8010 Graz | Austria
Visit us at www.innophore.com


Licence and Disclaimer

Privacy Policy

Copyright © 2022 Innophore.