Central Limit Free Energy Perturbation (CL-FEP): a tool for accurate end–state free energy estimations

Binding free energy calculations are essential for the study of chemical and biological systems. However, accurate methodologies in the field are often computationally demanding, which limits their applicability. As an alternative, we introduced the Central Limit Free Energy Perturbation (CL-FEP) approach. Without computing intermediate points of a system transformation, CL-FEP can deliver accurate free energy change values. In CL-FEP, the exponential average estimator is evaluated upon transforming the energy output of molecular dynamics simulations by applying the Central Limit theorem. Importantly, energies from explicit solvent simulations can be used to evaluate the estimator and no fitted parameters are introduced in our implementation. 
https://clfep.cbe.bci.tu-dortmund.de/

Multi-Scale Computational Approaches: QM/MM/CG implementation

Hybrid methods, typically Quantum Mechanics/Molecular Mechanics approaches (QM/MM), are at the core of our research. There is considerable interest in the development of coarse-grained (CG) force fields to improve the performance in MD simulations and geometry optimizations. We implemented a triple-layer Quantum Mechanics/Molecular Mechanics/Coarse Grained (QM/MM/CG) modelling approach to reduce computing times and extend the applicability of QM/MM calculations. We applied our implementation to the study of the effect of CG water on two enzymes: chorismate mutase and p-hydroxybenzoate hydroxylase.

PPI-Detect: a tool for predicting the interaction likelihood of protein–protein and protein–peptide pairs

Machine learning methods allow exploiting the information content in protein datasets. We introduced a procedure for the general-purpose numerical codification of polypeptides. Using this codification, we developed a support vector machine model (PPI-Detect), which allows predicting whether two proteins will interact or not. PPI-Detect outperformed other state of the art sequence-based predictors of protein–protein interactions.

The PPI-Detect web server is available at the following link:

https://protdcal-suite.cbe.bci.tu-dortmund.de/ppi_detect/

PPI-Affinity: a tool for predicting the binding affinity of protein–protein and protein–peptide complexes and for protein engineering applications

PPI-Affinity is a machine-learning tool developed to predict the binding affinity of protein–protein and protein–peptide complexes, based on their three-dimensional structure. We implemented PPI-Affinity as a web tool with three modules: (1)  for estimating the binding affinity of protein – protein and protein – peptide complexes, (2) for estimating the affinity of specified mutants and as a (3) protein-engineering module for generating mutant complexes and estimating their affinity. PPI-Affinity thus facilitates the virtual screening of protein–protein and protein–peptide interactions, with direct applications for hit identification and optimization in drug design projects as well as for protein engineering.

The PPI-Affinity web server is available at the following link:

https://protdcal-suite.cbe.bci.tu-dortmund.de/PPIAffinity/

ABP-Finder: a machine-learning tool for the identification of antibacterial peptides

To address the problem of multi-drug resistance in bacteria, we developed ABP-Finder, a tool specifically designed to identify antibacterial peptides (ABPs) and for the estimation of which type of bacteria is targeted by these peptides. ABP-Finder ranks within the top state-of-the-art predictors of antibacterial peptides, particularly in terms of precision. We also applied ABP-Finder for the identification of an antibacterial peptide by screening the human urine peptidome.

The ABP-Finder web server is available at the following link:

https://protdcal-suite.cbe.bci.tu-dortmund.de/ABP-Finder/

ProtDCal-Suite: a web server for studying proteins

We also introduced ProtDCal-Suite, a web server comprising a set of tools for studying proteins. The main module of ProtDCal-Suite is a software named ProtDCal (Ruiz-Blanco, Y.B., Paz, W., Green, J. et al. BMC Bioinformatics 16, 162 (2015). https://doi.org/10.1186/s12859-015-0586-0).
ProtDCal encodes the structural information of proteins in machine-learning-friendly vectors. Secondary modules of ProtDCal-Suite are tools developed for protein analysis that benefit from ProtDCal’s descriptors, including PPI-Detect, PPI-Affinity and ABP-Finder.

The ProtDcal suite web server is available at the following link:

https://protdcal-suite.cbe.bci.tu-dortmund.de/