SMOL Drug Design

AI powered drug discovery platform for faster and cost-effective de novo small molecule design

Context

Time, effort, and cost intensive drug discovery process

(Upper) Chemical structures of small-molecule drugs (SMOLs), i.e. naltrexone, methotrexate, and doxorubicin. (Lower) Space-filling models of the SMDs: carbon (dark grey), hydrogen (light grey), nitrogen (blue), oxygen (red).
Structures and Models of Small Molecule Drugs
Discovery of New Chemical Entities (NCE) with the desired biological activity is the foundational step for new therapeutics especially as the druggable target protein space is expanding. The vast chemical space of >1065 drug-like molecules requires sophisticated algorithms to assist scientific exploration.Advances in cheminformatics and computational chemistry combined with AI/ML are enabling chemists to have a more informed exploration of the physicochemical properties of the “beyond the rule of 5” (bRo5) chemical space. Avoidable Experiment Expenditure (AEE), estimated annually to be about $48B, is a serious concern for the bio-pharma industry. In silico techniques can significantly help reduce these costs. Drug candidates failing in late-stage clinical trials is a huge contributor to AEE.Therefore, discovery chemists need to factor in aspects such as ADMET properties, synthesizability etc. along with biological activity to improve the quality of leads generated for eventually improving the chances of success in clinical trials.
Our Solution

Highly repeatable and accelerated in silico pipeline to generate de novo molecular leads for a given target protein

Our solution enables an efficient and cost-effective in silico drug design to generate novel molecular leads for a given target protein. We help reduce AEE for BioPharma by combining the concepts of cheminformatics & computational chemistry with advances in AI/ML, Cloud & DevOps. Our solutions complement the traditional lab-based approaches with in-silico solutions to accelerate drug discovery. Key features of our solution include:

  • Generative models to navigate the vast ‘drug-like’ chemical space while optimizing for biological activity, synthesizability, drug-like characteristics, ADMET properties
  • High throughput molecular docking to estimate binding affinity of lead candidates. Top docked complexes are further explored for potential interactions that reveal novel binding mechanisms through a curated binding site analysis
  • Accelerated DFT computational pipelines to estimate DFT descriptors that improve ADMET predictions

The generative models can be customized for target proteins and molecular properties of interest. High throughput docking pipelines can be deployed on cloud-based Kubernetes environments with schedulers such as SLURM for auto-scaling and workload management.

SMOL drug design pipeline steps: given a target protein the steps include Ligand Generator, Ligand Evaluator, ‘Drug Likeliness and Toxicity Filters, Automated High Throughput Docking and ADME characterization towards identifying Drug Leads for Wet Lab Assays
Drug Design Pipeline
Highlights

Key components & strengths

Complete pipeline with comprehensive capabilities

Generating synthesizable drug-like molecules to distilling promising candidates as drug seeds for downstream wet lab assays

State of the art AI/ML tools & techniques

Generative models combined with Reinforcement learning, high-throughput docking evaluations and DFT descriptors based ADMET predictions

Customizable and configurable components

Modules built leveraging open-source packages,customizable pipeline with flexibility to choose and add specific components

On-demand cloud computing infrastructure

HPC clusters, cloud-based Kubernetes environment for auto-scaling, and schedulers such as SLURM for workload management, resulting in high throughput
Outcomes

Accelerate discovery of de novo SMOL drugs

Faster and Cost-effective

Identification of synthesizable drug-like molecules with affinity to target protein and promising ADMET properties using advanced machine learning techniques and high throughput docking pipelines

Low-effort Customization

With modular and configurable solution components (ADMET prediction, Retrosynthesis, Reaction Modeling, Virtual Screening, MD simulations etc.) across the molecular lead generation pipeline for different target proteins

Configurable and Scalable

Scalable computational resources with on-demand cloud-based High Performance Computing (HPC) clusters, infrastructure as a code (IaaC) approach and workload management techniques

Discover our offerings across the biopharma value chain

Learn more about our SMOL Drug Design