Software packages

Click image to access repository Software

Griffin - grid-based implicit force field

Griffin precalculates a potential force field for the molecules that the user selects for implicit treatment. Implicit means that these molecules are held in a fixed conformation, while the explicit molecules can move freely in the implicit force field. Griffin maps Coulomb and van der Waals interactions as well as surface forces, an efficient way for resolving clashes, onto a grid. For each grid point, Griffin calculates the force experienced by a dummy atom at that position from the entire implicit molecular system. In its initial version, Griffin is linked to a molecular dynamics simulation of the explicit molecules. Benefit of this approach is the abillty to resolve clashes even in highly complex topologies. In addition, the implicit force field approach implies a significant speedup.
Currently, we extend Griffin to perform rigid body dynamics of an arbitrary number of copies of the implicit molecules. This can be run both standalone as well as in conjunction with explicit molecular dynamics.

SmoothT– constructing low energy pathways from simulations

SmoothT is a a software for the construction and visualization of transitional pathways between conformational states. The input is an ensemble of molecular conformations in PDB format, along with a list of energy values associated with the conformations. Energies may be physical, statistical or heuristic estimates and combinations thereof. In addition, the user must specify starting and ending conformation, as well as a threshold for the chosen similarity measure, which can be either RMSD or a distance matrix score that is more sensitive to local similarities.
Although there are no inherent constraints on the source of the conformations, we describe two common example applications, one in the context of MD simulations, the second in the context of Monte Carlo (MC) algorithms such as docking and/or folding. To increase confidence, multiple sets of MD simulations are generally run at slightly different starting velocities. The simulations themselves may remain in the range of repetitive local oscillations over long periods of time. SmoothT can be applied in this context to derive a low energy-path combining multiple independent trajectories.
In MC algorithms such as Rosetta, a common strategy is to perform many short independent runs, resulting in a large ensemble where each pose represents a local minimum. Such ensembles can be repurposed and used as a starting point for modeling transitions. Combinations of MD and MC are also possible as long as they are evaluated with identical scoring schemes or potentials.
SmoothT performs two steps: First, a graph is created in which each conformation is represented by a node. All pairs of nodes that have a similarity score above the defined threshold are connected with an edge. In the second step, SmoothT determines the path through the graph that has the lowest energy barrier and also has the lowest integral over energy.
The output is a concatenated PDB file that can be used directly for interactive movie-like visualization with common molecular visualization software such as VMD. The output can also serve as a starting point to an enhanced sampling MD strategy, where each node would serve as a starting point for an individual simulation.
A strength of this fairly simple approach is that no assumptions are made about location or shape of the transition pathway.
Example (scroll down and press start button).


The initial activation step of T-cell based immune response pathways of the adaptive immune system is the binding of the T-cell receptor (TCR) to the epitope-presenting major histocompatibility complex (MHC). Differences in MHC between individuals are essential for the species survival, underscoring the importance of this branch of the immune system. Since TCR-epitope-MHC structures have a high degree of conservation, threading is a rapid and accurate method to determine potential activation of the adaptive immune system. While MHC profiling is a well established standard, sequencing of the TCR repertoire has also been introduced in recent years. This will allow to predict whether an epitope is a candidate for a vaccine.

The EpitopeThreader is fusing amino acid side-chains onto Cβ positions. In order to change the sequence on MHC, epitope or TCR only the identity and thus the potentials have to be exchanged. The CHARMM force is applied for the coarse-grained energy calculation. Millions of sequence combinations can be evaluated within seconds.


Cryo electron microscopy (EM) has made tremendous progress in the last decade. Images can be made of very large systems, even entire viruses, sometimes at quite high resolution. Much more often, however, medium resolutions are obtained. Resolutions that do not allow to derive a model in atomistic detail directly.
bcl::EM-fold is a Monte-Carlo algorithm that places secondary structure elements in densities. bcl::EM-fold incorporates the densities as potentials which allows to combine it with simple heuristic scoring schemes. A publication about bcl::EM-fold was recommended by the 'Faculty of 1000'.


AlignMe generalizes the concept of similarity by considering evolutionary and biochemical properties. Furthermore, expert knowledge and experimental data can be included as anchors.

Protein Prompt

ProteinPrompt predicts protein-protein interractions (PPIs) from sequence. We provide two implementations, one using Random Forests, the other Graph Neural Networks.