Contact Us
By submitting request you agree to our Privacy Policy
AI-Powered Drug Discovery for Bcr-Abl Inhibitors
Client: Pharmaceutical research organization
QA
Expertise:
Tech Stack:
DevOps
UX/UI
PM
AI/ML
Frontend
Seaborn
Matplotlib
Python
Google Cloud
AutoDock
PyMOL
TensorFlow
PyTorch
NumPy
AWS
RDKit
Business Analyst
Backend
The client is a pharmaceutical research organization focused on developing advanced therapeutic solutions for Chronic Myeloid Leukemia (CML). Their objective is to utilize AI and ML technologies to overcome the challenges of drug resistance and toxicity in existing treatments.
About the Client:
The goal was to design a generative hetero-encoder model capable of creating novel Bcr-Abl tyrosine kinase inhibitors with reduced toxicity and improved resistance to mutations, such as the T315I mutation. The project employed advanced molecular docking and machine learning techniques to streamline the drug discovery process and identify new potential compounds.
About the Project
Challenges & Objectives
Challenges:
- Toxicity and Resistance: Existing inhibitors like imatinib and ponatinib show high toxicity and resistance due to the T315I mutation.
- Complexity of Drug Discovery: Molecular docking and compound screening require significant computational resources and time.
- Accuracy in Prediction: Ensuring that newly designed compounds meet pharmacological and binding efficacy standards.
- Develop a deep generative hetero-encoder model for de novo compound design.
- Optimize molecular docking to improve the efficiency of filtering and analyzing generated compounds.
- Identify compounds with high binding affinity and pharmacological potential for further experimental validation
Implementation
The development process followed an iterative approach to achieve optimal performance and user experience.
01
Data Collection and Training Library Assembly:
- Selected 120,000 compounds containing the aryl-aminopyrimidine fragment from DrugBank.
- Converted chemical structures to SMILES formats and prepared molecular descriptors.
02
Molecular Docking Preparation:
- Generated molecular docking complexes targeting the ATP-binding site of wild-type and T315I-mutated Bcr-Abl kinase.
- Calculated binding free energy values for compound evaluation.
03
Model Construction:
- Developed a hetero-encoder architecture to process molecular input in multiple formats.
- Integrated energy calculation models to improve prediction accuracy.
04
Molecule Generation:
- Trained the neural network using the molecular library and validated its performance.
- Designed new small molecules targeting Bcr-Abl kinase and validated results through docking and molecular dynamics simulations.
05
Further Research and Analysis:
- Identified top-scoring compounds with high binding affinity and pharmacological properties.
- Compared results to established drugs like imatinib and ponatinib to ensure comparable efficacy.
Outcomes & Business Impact
- New Compounds: Identified five promising compounds with strong binding affinity and reduced toxicity.
- Broad-Spectrum Potential: Created inhibitors capable of targeting wild-type and T315I-mutated Bcr-Abl kinase.
- Efficient Drug Discovery: Computational methods accelerated the process, cutting time and costs.
- Improved Models: Developed an innovative hetero-encoder model for inhibitor design, enhancing prediction accuracy.
- Cost Reduction: Streamlined computational workflows reduced drug discovery costs by 40%.
- Accelerated Development: Shortened the compound design timeline by 35%.
- High-Quality Outcomes: Identified compounds with potential for further experimental validation, paving the way for next-generation cancer therapies.
- 5 Promising Compounds: Displayed strong binding affinity and favorable pharmacological profiles.
- 40% Cost Reduction: Lowered computational and experimental expenses.
- 35% Time Savings: Accelerated the drug discovery process through AI-driven modeling.
Results & ROI
Let us help you with your business challenges
Contact us to schedule a call or set up a meeting
