Exscalate Logo
Exscalate Platform

The power of supercomputing

Supercomputers have seen incredible advancement in recent years. Machines thousands of times more powerful than desktop computers are being leveraged to accelerate the drug discovery process. This supercomputing power together with other advanced technologies, such as AI and big data analytics, enables Exscalate to rapidly and effectively address the complexity of a flexible target. Our platform is able to evaluate two trillion molecules at a rate of up to 60 million per second for each new target.
How we do it 

The Exscalate platform is built upon three key building blocks:

Tangible Chemical Space (TCS)

A tangible chemical space (TCS) consisting of the world’s largest virtual ligand library of over 2 trillion organic chemistry molecules. This includes ready to screen annotated libraries, such as safe-in-man drugs (>5K), natural products (>5K), oligopeptides (>5K) as well as ad hoc libraries of allosteric modulators.  The TCS also includes all the information required to easily synthesize each of these molecules in a one-step reaction. With the power of AI, maximum knowledge can be extracted for structure-activity relationships.

Comprehensive Therapeutic Target Database (CTTD)

A comprehensive therapeutic target database (CTTD) combines all known annotated druggable targets with decades of research on protein–ligand and protein–protein interactions. This database leverages advanced annotations of more than 45,000 binding sites (mainly allosteric) of small molecules and peptides generated by the analysis of large-scale molecular dynamics simulations combined with a proprietary tool for the analysis of the targets’ pocketome [ref]. CTTD is enriched with proprietary protein functional annotation.

Ligand Generator (LiGen)

LiGen (Ligand Generator) is a powerful in silico simulator that runs on High Performance Computing (HPC) architectures. The binding of 16.5 billion molecules can be virtually tested on a target protein — including many conformations of the same protein — in just one hour. These virtual simulations increase the quality of expected delivery, reduce time and ensure transactional clinical success.


The tool allows prediction of possible toxicity in 8 models related to the main
toxicities with which drugs must be compared: cardio, neuro, hepato. The tool is able to predict more than 600 drug activities and predict associated risk profiles. This tool is critical to remove molecules with other
probabilities of being toxic to humans from the development process and focus
on safe ones right away. 


Our scoring function has a higher accuracy rate:

more accurate than GLIDE
more accurate than PLANTS

When using the PDBind 2020 dataset

CPU Performance

LIGEN is capabile running on both CPU or GPU

On CPU LIGEN perform 30 molecular docking simulations per second
This is 600x faster than the GLIDE (CPU only)
This is 230x faster than the PLANTS (CPU only)
GPU Performance

GPU based on the NVIDIA Ampere architecture

k mol/s
On GPU LIGEN perform 6000 molecular docking simulations per second
This is 120,000 times faster than the GLIDE CPU approach
This is 46,000 faster than the PLANTS CPU approach

GPU Performance stats based on 4x A100 TENSOR CORE infrastructure

We obtained these results thanks to a close collaboration with NVIDIA. Read More

The workflow

The process of using the platform

1) Disease Poly Pharmacology Phase

Utilizing the EXSCALATE Knowledge Graph, we identify the most important targets.

2) Disease Modifying Target phase

Utilizing the information in our vast database (CTTD), we find and further modify the targets by placing them in the context of the pathology.

3) Lead design and Safety Profiling Phase

Once we know the pathology and the characteristics of the possible modulator, we use the TCS tool to define the most suitable chemical space. Thanks to the LIGEN virtual screening platform and PROFHEx liability profiler, we are able to design ligand targeted libraries with the highest in silico safety profile. Available for experimental validation.

4) Experimental Preclinical Validation Phase

The active compounds identified in the previous step are then validated in phenotypic screening and then tested for efficacy in preclinical and translational models.

5) Lead Optimization phase

The validated information generated in the previous phases is used to repeat the process. The simulation is rerun in LiGen and together with the learning capabilities of AI, the platform predicts a new optimized molecule.

6) Preclinical Tox phase

Any optimized molecules that represents the highest level of accuracy of chemistry is progressed to preclinical toxicology studies that analyze the safety effects of the drugs.

7) Investigational new drug (IND) phase

Shown to be safe, the molecule can now progress into clinical development.