CHEESE: AI-Based Tools for Accelerated Drug Discovery
The CHEESE Platform offers several key tools for (not only) medicinal and computational chemists:
CHEESE Search, CHEESE Modeller, CHEESE Electrostatics, and CHEESE Explorer.
CHEESE Platform Availability
CHEESE tools are available in a free limited demo version and in premium version, in dedicated and on premise deployment that provide additional functionality and safety/privacy features.
You can contact us for more info about CHEESE products at info@deepmedchem.com.
CHEESE Search (Chemical Embeddings Search Engine) is set to revolutionise drug discovery and chemical informatics by enabling rapid and efficient ligand-based similarity searches of large molecular databases and spaces. Traditional methods are often slow and impractical, but CHEESE overcomes these challenges with its fast and scalable approach.
CHEESE Search provides very straightforward UI to make its users efficient. In addition, CHEESE Search API allows you to make it part of your existing workflows. In comparison with existing molecular search tools it provides functionality for more precise search work with the molecular databases.
CHEESE learns and represents the molecular space, making it the core tool for molecule similarity searches. It supports various molecular similarity metrics, including 2D Fingerprints, 3D Shape, and 3D Electrostatic. In the advanced versions CHEESE Search is able to make your molecular databases searchable.
CHEESE Search is available at cheese.deepmedchem.com.
CHEESE Modeller is designed to simplify the process of building predictive models from your experimental data even when your AI or machine learning knowledge is limited. Here’s how it can help:
- User-Friendly Interface: CHEESE Modeller provides an intuitive GUI, making it accessible even for users with minimal AI expertise.
- Rapid Model Building: You can build a predictive model from your dataset in seconds to a few minutes. This includes estimation of the predictor quality on your dataset.
- Efficient Predictions: Run predictions on any large database quickly. In our customer case study, the complete cycle—from experimental data to ranking 40M molecules—took just one hour.
CHEESE Modeller empowers you to harness the power of AI for drug development, streamlining the process and enhancing the quality and diversity of your candidate structures.
CHEESE Electrostatics is a cutting-edge tool designed to rapidly and accurately estimate Electrostatic Potential (ESP) and Restrained Electrostatic Potential (RESP) for molecules involved in drug development. This tool provides a significant speed advantage, accomplishing what typically takes hours, days, or even weeks with Density Functional Theory (DFT) in just seconds, while maintaining a high correlation (0.98) with DFT calculations.
Key Features:
- Fast ESP/RESP Estimations: Achieve results in seconds.
- High Accuracy: Comparable to DFT with a 0.98 correlation.
- User-Friendly Interface: Designed for ease of use by both novices and professionals.
- Multiple Charge Models: Includes Gasteiger and MMFF charges for comprehensive comparisons.
Visit CHEESE Electrostatics to use the tool. No installation necessary as the tool operates entirely online.
The CHEESE Explorer is an AI-based molecular space exploration tool designed to generate and visualise chemical representations based on the trained latent space and is based on shape or electrostatic similarity of molecules.
This tool offers a fast and efficient way to explore molecular datasets, making it an invaluable asset for researchers in the field of medicinal chemistry and drug discovery.
Key Features:
Chemical Representations:
- CHEESE Explorer generates chemical representations swiftly, focusing on shape or electrostatic similarity.
Dimensionality Reduction:
- The generated representations can be plotted using dimensionality reduction techniques, enabling users to visualise the dataset effectively.
Visualisation and Exploration:
- Users can color the dataset by clusters or specific properties.
- The tool allows for the exploration of data points (molecules) that are close to each other in the latent space.
Interpretable Latent Space:
- CHEESE Explorer ensures that the distances between data points in the latent space correlate highly with their real similarities.
-
Unlike other AI models, CHEESE Explorer guarantees the interpretability of its latent space. For example:
- Molecules with a 0.9 cosine similarity or a 0.1 Euclidean distance have proportionally the same fingerprint, shape, or electrostatic similarity, depending on the model.
Clustering and Similarity-Network Computation:
- The tool supports subsequent clustering or similarity-network computation, facilitating deeper analysis and insights.
News
Deep MedChem was presenting CHEESE at BIO'24 in San Diego (June 2024)
It was a privilege for Deep MedChem to be part the Czech Pavilion at BIO'24 together with great companies and institutions. The support from the Czech Ministry of Industry and Trade was excellent and made our presence the more enjoyable.
Fours days full of meetings, interactions and interesting discussions were a time very well spent!
Enamine REAL is now supported in CHEESE (May 2024)
We are excited to announce that Enamine-REAL DB is now covered by CHEESE Search—our super fast AI-based similarity search engine.
We now offer two databases to retrieve similar molecules from by CHEESE Search: ZINC (700M molecules) and Enamine Ltd.'s Enamine-REAL (5.5B molecules). We are able to index basically any molecular database now and make it part of CHEESE Search.
We improved the speed and accuracy of the search even further—you get the results in sub-second times.
CHEESE Electrostatics in open beta NOW! (May 2024)
We were pleased with the interest from the expert community after we announced the results of our CHEESE Electrostatics tool. We promised to make it available after some time and the time is NOW!
You can try the CHEESE Electrostatics in beta yourself.
CHEESE Talk in Miton AI Talks series (March 2024)
Our CTO Miroslav Lžičař gave a talk in the Miton AI times series on how we make billion- and trillion-scale molecular spaces accessible for search and filtering by properties. You can watch the recording on YouTube:
CHEESE Modeller is available for customers! (February 2024)
Build your custom fast predictive models for molecular properties with CHEESE Modeller with state of the art prediction performance.
Key benefits of CHEESE Modeller: speed of training of the predictive models, the speed of prediction by the models, very good performance of the predictive models (state of the art in most cases of ADMET properties), ability to train the models on your own data and (optionally) on your dedicated instance (or even on AWS or on premises on your own servers). Contact us!
On-Premise Version of CHEESE Search Now Available! (January 2024)
Due to popular demand our team has dedicated substantial efforts to bring you the on-premise version of CHEESE Search! We understand that many customers prioritise the security of their sensitive data and prefer not to transmit it over the internet. With the introduction of the OnPrem CHEESE, you can now experience all the incredible benefits of CHEESE while maintaining the confidentiality of your data. Discover more details in our documentation section titled "OnPrem CHEESE."
Deep MedChem will comercialize CHEESE family of products (December 2023)
We are thrilled to announce the establishment of Deep MedChem, a cutting-edge company dedicated to commercialise the CHEESE family of products—advanced molecular search tools and their extensions to the market. Founded in the first half of December, Deep MedChem is committed to revolutionising the landscape of drug development AI tools and their accessibility.
CHEESE Webinar with ChemSpace (August 10, 2023)
We had the pleasure to present details of CHEESE in the webinar series of ChemSpace.
Presentation slides from the webinar:
New Content on CHEESE-DOCS (July 27, 2023)
- Look on Prototypes page to see what we are working on. And Benchmarks page to see how CHEESE performs in various tasks. Changelog page contains information about new features in each version of CHEESE webpage and features of upcoming versions.