Periodic Reporting for period 2 - ATLAS (Next generation HPC/QC in silico drug discovery platform for de novo design)
Berichtszeitraum: 2023-09-01 bis 2024-08-31
Drug Discovery is a long & costly process currently mostly driven by wet lab experiments. The reason, simulations tools did not achieve the same reliability and performance than in the aerospace, automotive or engineering sectors. This is exactly what the Atlas project aims to change: developing a predictive, reliable and cost-effective platform to accelerate the drug discovery process.
ATLAS is also the name of Qubit’s ultra-high resolution drug design platform. It enables drug hunters to discover, optimise & validate drug candidates and carry out complex computations at scale, on any hardware, whether on-premises or on the cloud. It leverages the cutting-edge scientific developments of our scientific founders & team in the field of molecular dynamics and more broadly in in-silico drug discovery. The platform allows tackling, CPU, GPU and QPU (Quantum Processing Units) to benefit from most advanced High Performance Computing (HPC) infrastructures and technologies to provide our drug discoverer a reliable, resilient, performant, scalable and secured tool.
The main use cases are:
- Provide great and secured user experience:
We make sure the provided user experience is a “must have” (and not a “nice to have”) is critical to ensure full users’ commitment to the use of the platform.
- Design advanced Drug Discovery (DD) pipelines: The main purpose of Drug Discovery is to create active compounds molecules. The capacity of designing efficient pipelines with cutting edge technology is central.
- Provide state of the art molecular visualisation and analysis tools: At Qubit Pharmaceuticals, we believe humans play a central role in Drug Discovery. The platform is at the service of the Drug Designer. The platform must provide the best tools to analyse and visualise molecules.
- Help and support users: Drug designers shall only focus on Drug Discovery and Drug Design. It is central to help and support them to take advantage of the best of the technology.
- Scale on demand
Drug Discovery is inherently a large-scale problem. It requires the ability to deal with very large datasets (i.e finding the best compound in a chemical space of 1060), to perform complex algorithmics (large matrix inversions) at scale & speed… hence HPC and involves large team with various expertise. This is the reason why scalability of our infrastructure is at the heart of the platform design.
- Monitor and report performances and costs:
We share a continuous improvement mindset, monitoring, tracking, reporting, alerting on the state of the platform is crucial to be as proactive as possible.
- Manage a database of compounds: Compounds are the Gold of Drug Discovery, a reliable and secured database with highly qualified data is a “must have” for Qubit Pharmaceuticals
ATLAS is also the name of Qubit’s ultra-high resolution drug design platform. It enables drug hunters to discover, optimise & validate drug candidates and carry out complex computations at scale, on any hardware, whether on-premises or on the cloud. It leverages the cutting-edge scientific developments of our scientific founders & team in the field of molecular dynamics and more broadly in in-silico drug discovery. The platform allows tackling, CPU, GPU and QPU (Quantum Processing Units) to benefit from most advanced High Performance Computing (HPC) infrastructures and technologies to provide our drug discoverer a reliable, resilient, performant, scalable and secured tool.
The main use cases are:
- Provide great and secured user experience:
We make sure the provided user experience is a “must have” (and not a “nice to have”) is critical to ensure full users’ commitment to the use of the platform.
- Design advanced Drug Discovery (DD) pipelines: The main purpose of Drug Discovery is to create active compounds molecules. The capacity of designing efficient pipelines with cutting edge technology is central.
- Provide state of the art molecular visualisation and analysis tools: At Qubit Pharmaceuticals, we believe humans play a central role in Drug Discovery. The platform is at the service of the Drug Designer. The platform must provide the best tools to analyse and visualise molecules.
- Help and support users: Drug designers shall only focus on Drug Discovery and Drug Design. It is central to help and support them to take advantage of the best of the technology.
- Scale on demand
Drug Discovery is inherently a large-scale problem. It requires the ability to deal with very large datasets (i.e finding the best compound in a chemical space of 1060), to perform complex algorithmics (large matrix inversions) at scale & speed… hence HPC and involves large team with various expertise. This is the reason why scalability of our infrastructure is at the heart of the platform design.
- Monitor and report performances and costs:
We share a continuous improvement mindset, monitoring, tracking, reporting, alerting on the state of the platform is crucial to be as proactive as possible.
- Manage a database of compounds: Compounds are the Gold of Drug Discovery, a reliable and secured database with highly qualified data is a “must have” for Qubit Pharmaceuticals
We test the results of our R&D efforts with end-users in real drug discovery projects: we have launched 6 different DD projects covering oncology and inflammation, throughout very hard targets such as metalloenzyme or GPCR.
Globally reduce time and cost to solution (and therefore associated cost) to obtain accurate ABFEB values on complex systems using new generation accurate polarizable force fields (PFF). We developed beyond state of the art technologies such as lambda-ABFE to speed up computation and reduce costs. In parallel Atlas provides a super agile means to develop optimised and scalable digital protocols to achieve industrially qualified results. In addition, a large AI-ML pipeline has been developed to enhance and better leverage the acquired data. Huge work has been done to ensure data quality in our fully functional database of several tens of millions of molecules.
The drug discovery pipeline consists in exploring the chemical space for a compound showing a proper activity & profile with a protein of interest. This requires a "funnel" like workflow: starting from an extremely large number of molecules and ending with a few effective ones. Atlas allows automating the steps and provides most accurate results BEFORE ABFE calculation. ABFE calculation shall be reserved to ligands that can already demonstrate to be lead candidates ready for pre-clinical trials.
We have integrated QPUs in our drug discovery pipeline to exploit near-term quantum advantages. This has been done both on emulators (PASQAL, Bracket, Qiskit) and real machines (PASAL, AWS Bracket). Our Quantum Computation strategy has grown to become the central focus of the R&D team by developing Hyperion: the quantum simulator for molecular models.
Globally reduce time and cost to solution (and therefore associated cost) to obtain accurate ABFEB values on complex systems using new generation accurate polarizable force fields (PFF). We developed beyond state of the art technologies such as lambda-ABFE to speed up computation and reduce costs. In parallel Atlas provides a super agile means to develop optimised and scalable digital protocols to achieve industrially qualified results. In addition, a large AI-ML pipeline has been developed to enhance and better leverage the acquired data. Huge work has been done to ensure data quality in our fully functional database of several tens of millions of molecules.
The drug discovery pipeline consists in exploring the chemical space for a compound showing a proper activity & profile with a protein of interest. This requires a "funnel" like workflow: starting from an extremely large number of molecules and ending with a few effective ones. Atlas allows automating the steps and provides most accurate results BEFORE ABFE calculation. ABFE calculation shall be reserved to ligands that can already demonstrate to be lead candidates ready for pre-clinical trials.
We have integrated QPUs in our drug discovery pipeline to exploit near-term quantum advantages. This has been done both on emulators (PASQAL, Bracket, Qiskit) and real machines (PASAL, AWS Bracket). Our Quantum Computation strategy has grown to become the central focus of the R&D team by developing Hyperion: the quantum simulator for molecular models.
The results of this last year of project are listed as follows.
- 6 publications and 3 patents (over the entire program we have issued 12 scientific papers and 8 patents)
- synthesise of preclinical leads on multiple chemical series competing with Pfizer first in class compound on KAT6A.
- Qubit-Pharmaceuticals has been awarded by the i-Demo programme with €8 million funding from the France 2030 plan for its quantum simulator Hyperion-1..
- We have decreased by 4 the cost of ABFE computation from around €400 down to €100 per compound in one year.
- We have integrated a complete ADMETox ML - based predictor with up to 92% accuracy compared to experimental data.
- 6 publications and 3 patents (over the entire program we have issued 12 scientific papers and 8 patents)
- synthesise of preclinical leads on multiple chemical series competing with Pfizer first in class compound on KAT6A.
- Qubit-Pharmaceuticals has been awarded by the i-Demo programme with €8 million funding from the France 2030 plan for its quantum simulator Hyperion-1..
- We have decreased by 4 the cost of ABFE computation from around €400 down to €100 per compound in one year.
- We have integrated a complete ADMETox ML - based predictor with up to 92% accuracy compared to experimental data.