Molecular Discovery Services

Tailored solutions for your drug discovery projects

How does it work?

We provide an array of services to solve the most challenging computer-based drug discovery problems. These services are based on molecular dynamics and machine learning, Acellera's two fields of expertise. All computations are run on Acellera infrastructure and the IP generated during the execution of a service (such as the results of the simulations) belongs to the client.

Your data will only be accessible to the team of employees working on the project. Transfer of data between the client and Acellera Ltd. is done through channels that comply with the highest standards of security. Acellera has collaborated with top pharmaceutical companies and has an extensive scientific production in the fields of molecular dynamics and machine learning applied to drug discovery.

Molecular Dynamics Simulations

Atomic scale studies for your system of choice.

These are some of the services that we offer. If you are not sure about whether your project fits these examples, please feel free to contact us, and let us know more about the specifics of your project.

  • IN SILICO BINDING ASSAY (ISBA): Using molecular dynamics simulations of small molecules or fragments, we can obtain accurate predictions of binding poses and affinities for protein-ligand pairs. The two videos on the right column showcase some examples, as well as this recent article from Acellera. We invite you to visit the Science section of our page to see more articles covering this topic.

  • PROTEIN CONFORMATION STUDY: Proteins are not a static macromolecule. If we let them evolve over time using MD, we can sample different conformations, and reach, for example, the active states of a GPCR starting from its inactive conformation.

  • CRYPTIC POCKET ANALYSIS WITH CRYPTICSCOUT: We can determine binding hotspots (catalytic, allosteric and hidden sites) on your target using mixed-solvent molecular dynamics simulations in presence of small probes such as benzene, isopropanol or imidazol. Cryptiscout scans the protein surface with these small probes to predict cryptic sites and druggable cavities.

  • GPU-ENHANCED: Molecular Dynamics simulations were of limited value a few years ago, until GPUs became mainstream. Now, using our cluster, we can simulate microseconds of complex systems in a matter of weeks. Together with the improvements in force fields, MD can offer accurate approximations of real molecular behavior, and has proven useful in different stages of drug development, as well as in understanding the biochemistry basis of physiological events. Our team of expert scientists can provide up to a millisecond of simulation time for your system of choice, as well as analyze these results, and provide you with structural insights of your target.

Machine Learning

Deep Learning and Generative models applied to virtual screening and hit discovery.

  • REPURPOSING, DEORPHANING AND OFF TARGET PREDICTION WITH PATHWAYMAP: Fast prediction of the interaction between a set of ligands and major human biological and signaling pathways using state-of-the-art self-normalizing neural networks. Main use for repurposing, deorphaning and off-target prediction.

  • PREDICTION OF RELATIVE AFFINITY WITH DELTADELTA: Given a set of molecules sampled from a congeneric series with known binding affinity to a target, a neural network will be trained to predict the binding affinity of any other molecule sampled from that same congeneric series.

  • DISCRIMINATION OF BINDERS FROM DECOYS WITH BINDSCOPE: We perform the re-ranking of a structural virtual screening of a library of compounds against your protein of interest using a neural-network-based predictor of binding. While ideally the HTS should be provided by the costumer, we can optionally run the virtual screening using our in-house docking tools. The algorithm will classify the ligands and rank them according to their calculated probability (from 0 to 1) to bind the target.

  • PREDICTION OF THE BINDING AFFINITY OF YOUR LIGAND WITH KDEEP: Predict the absolute binding affinity (dG) of a set of ligands experimentally resolved or docked in a protein using state-of-the-art neural network-based predictor. Only known binders can be evaluated by KDeep (for instance, ligands after functional assays). This tool was used in two winning submissions of the D3R Grand challenge 4.

  • DE NOVO DESIGN OF LEADS MOLECULES WITH LIGDREAM: Generation of novel compounds from a seed ligand through generative, shape-based neural network decoding.

We invite you to visit our profile in Science Exchange:

alejandroDrug Discovery Service