Protein structure is a valuable tool for understanding protein function. Protein function is thought to be significantly influenced by protein dynamics. Therefore, to fully comprehend protein function at the molecular level, flexibility must now be taken into account in addition to structural study. Information on protein flexibility is essential for understanding key molecular mechanisms, such as protein stability, interactions with other molecules, and protein function in general.
In general, Alfa Chemistry relies on the backbone fluctuations in solution observed in molecular dynamics simulations of B-factors from X-ray experiments to analyze the local structural flexibility features in proteins. The need for computational techniques to predict protein flexibility from amino acid sequences is critical as the gap between the number of solved protein structures and the available protein sequences widens. We offer protein flexibility prediction services to our clients using MEDUSA, a deep learning-based protein flexibility prediction tool.
Fig 1. MEDUSA workflow. A. Example of MEDUSA input and the preprocessing step. B. Input format for a prediction. C. For each encoded sequence position flexibility class attribution is performed by a convolutional neural network of the given architecture. D. Example of the MEDUSA web-server output for three classes prediction. (Meersche Y, et al. 2021)
Our Computational Method
MEDUSA assigns a flexibility class to each place in a protein sequence by using evolutionary data gleaned from protein homologous sequences and the physicochemical characteristics of amino acids as input to a convolutional neural network. MEDUSA offers two-, three-, and five-level flexibility predictions after being trained on a non-redundant X-ray structural dataset. MEDUSA is available free of charge as a web server, providing clear visualization of prediction results, and as a stand-alone tool. Users can detect possibly highly distorted protein areas and the overall dynamic characteristics of proteins by analyzing MEDUSA data.
The MEDUSA web server provides users with a comprehensive visualization of the prediction results, which can be downloaded in their original format. For a better understanding of the dynamic characteristics of proteins with uncertain or disordered 3D structures, MEDUSA's information is crucial.
Fig 2. Performance of MEDUSA flexibility class predictions. A. Comparison of MEDUSA performance for two binary classification problems with the state-of-the-art utility PROFbval. B. Confusion matrices for predictions performed by MEDUSA and PROFbval for the binary classification problems. C. and D. Confusion matrices of MEDUSA performance for the three- and five-class prediction respectively. (Meersche Y, et al. 2021)
|Protein Flexibility Prediction
|We provide all raw data and analysis services to our customers.
|Our services require specific requirements from you.
|According to customers' needs
|Please contact us for an inquiry
A non-redundant X-ray structural dataset with resolution below 2.2 Å, R-value below 0.3, and sequence consistency below 25% was extracted from the Protein Data Bank using the PISCES web server. Only files with complete B-factor data were saved. There are 9880 proteins in the complete dataset.
Flexibility Class Definition
For each residue, MEDUSA forecasts the degree of flexibility of the protein's -carbon B-factor. The B-factor of each protein was standardized using the following equation in order to compare the B-factor values of various proteins acquired under various experimental settings:
Bnorm = (B - < B > )/std(B), where mean and standard deviation values were calculated across the considered protein residues.
The normalized B-factors are classified into categories in four different ways:
- 2 classes, Non-Strict case: Bthresh = −0.3.
- 2 classes, Strict case: Bthresh = 0.03.
- 3 classes: Bthresh = −0.5; 1.
- 5 classes: Bthresh = −1; 0; 1; 2.
Our protein flexibility prediction services significantly reduce costs, facilitate further experimentation, and accelerate the drug design process for our global customers. Our personalized, full-service approach will meet your innovative learning needs. If you are interested in our services, please feel free to contact us. We would be happy to work with you and see you succeed!
- Meersche Y, et al. (2021). "MEDUSA: Prediction of Protein Flexibility from Sequence." J Mol Biol. 433(11): 166882.