### Summary
Schrödinger, a drug discovery company, combines physics and machine learning to develop new drugs. The company collaborates with other drug developers and utilizes AI and physics-based principles to identify new drugs for various diseases.
### Facts
- Schrödinger sees traditional chemists, not other AI-based drug discovery companies, as its competition.
- The company combines AI with physics-based principles to identify new drugs and targets for various diseases.
- Schrödinger relies on machine learning to generate large amounts of data for training its models, as physics-based calculations are slow.
- Schrödinger's drug discovery efforts are supported by physics and machine learning.
- The company has over 1,750 customers for its drug discovery software and 13 active collaboration projects with biopharma partners.
- Schrödinger's partners include Bristol Myers Squibb, Eli Lilly, and Takeda Pharmaceutical.
- The company also has its own pipeline of 19 active programs, with the first candidate entering clinical trials in 2022.
- Schrödinger's candidates include inhibitors for MALT1, CDC7, and Wee1 genes, with plans to move more programs into clinical trials in the next decade.
- The company has a chief medical officer responsible for clinical development and regulatory strategy.
- Schrödinger reported net income of $4.3 million in Q2 2023 and lowered its guidance for drug discovery revenue, citing delays in milestone achievements by collaboration partners.
- The company's software revenue remained stable and Schrödinger raised its guidance for 2023.
🧪💻🧬🔬💊🔍
Artificial intelligence (AI) techniques, particularly machine learning, are increasingly being used in drug research and development (R&D), with applications expanding beyond small molecules to include large-molecule modalities such as antibodies, gene therapies, and RNA-based therapies. These therapies, which make up a significant portion of the biopharma industry's current and future commercial potential, are expected to represent approximately 50% of the oncology market by revenue in 2030, with the majority coming from antibodies.
A study from the University of South Australia has utilized machine learning to identify metabolic biomarkers that can predict the risk of cancer, offering the potential for early detection and prevention.
A review published in Engineering explores the potential of machine learning (ML) in revolutionizing chemical research, providing insights into popular ML algorithms and their applications in chemistry.
A research team at Pohang University of Science and Technology (POSTECH) has developed a machine learning approach to predict drug outcomes and side effects before clinical trials, using gene perturbation effects to evaluate the discrepancies between cells and humans.
Machine-learning algorithms developed at the University of Michigan can identify problem areas in antibodies that cause them to bind with non-target molecules, enabling researchers to modify the antibodies and optimize their effectiveness in fighting diseases like Parkinson's, Alzheimer's, and colorectal cancer.
Professor Brad Pentelute and the Pentelute Lab at MIT are using a combination of chemistry, biology, and engineering to develop new techniques and platforms that have the potential to revolutionize therapeutics, including using nature-inspired research to solve protein delivery problems and building an automated protein printing machine. They are also using machine learning and automation to discover new peptides and proteins that can be used in cancer treatment and other applications, with the goal of rapidly designing molecules with new functions and advancing the field of AI-driven molecule design.
The use of generative AI, combined with federated and active learning, can accelerate the development of protein drugs by improving predictions of drug properties and enabling collaboration among biopharmaceutical companies while protecting their competitive interests.
The use of AI tools like AlphaFold to predict protein structures and aid in drug discovery is gaining momentum, but questions remain about the quality and validation of the predicted interactions.
Researchers from the University of Eastern Finland, along with industry and supercomputers, have used machine learning to speed up virtual drug screening by 10-fold, reducing the processing time of 1.56 billion drug-like molecules.
Machine learning has been used to predict where drugs will concentrate in cellular compartments called condensates, which can affect the efficacy and safety of the drugs, and may lead to improved drug design.
