Dear colleagues and friends,
Welcome to the latest issue of the Deep MedChem Newsletter!
In recent weeks, we've seen great advancements that prove our commitment to providing unique AI-based tools for the drug discovery community: Firstly, we are excited to showcase the CHEESE Modeller through a detailed case study. Secondly, the launch of our CHEESE Electrostatics tool has garnered enthusiastic feedback. Let’s dive a little deeper:
We showed how AI can increase the speed of drug development in a practical pre-screening case study with our partner Bridgene Biosciences.
In this rather typical use case, our partner had a limited data set of structures with evaluated activity for a particular target and the goal was to utilize this data to screen the Enamine database (or more precisely a 40M subset of Enamine) for additional good candidate structures.
One option is to go through the DB and spend a long time doing so. OR you can build a predictive model based on the already measured dataset and filter the molecular database by this model to choose candidates of better quality.
When we did this with our partner pharma company we were able to increase the number of structures that were successfully validated in subsequent steps (docking to a target) 147 times!
Also the variability of clusters and scaffolds increased (3 times and 20 times respectively). The time to do the screening was also dramatically reduced—from 200 days to 25 hrs.
You may be wondering how do you build such a model with only little AI/Machine learning knowledge:
Well, CHEESE Modeller is the answer for you as it was for our partner. It allows you to take your data, build the predictive model in matter of seconds/short minutes, and run predictions on any large database very quickly—this full circle takes one hour to get from your experimental data to ranking of 40M molecules with respect to their properties (activity towards a target in our above example).
We are thrilled by the very positive response from the expert community following our announcement of the CHEESE Electrostatics tool preview. As promised, we are excited to announce that the tool is now available for all to try in its beta version!
Quick Recap of Benefits: Traditional methods like DFT can take days, or even weeks, to yield results, but with CHEESE Electrostatics, you can achieve similar accuracy (0.98 correlation with DFT) in just seconds. Explore the capabilities of CHEESE Electrostatics by visiting our website [here](https://electrostatics.deepmedchem.com/) and taking part in the beta testing. We eagerly anticipate your feedback, comments, and recommendations to further refine our service and enhance its value.
Additional Features: Alongside our innovative method for ESP/RESP estimation, we also provide comparisons with widely recognised Gasteiger charges and MMFF charges.
Licensing Update: We are currently offering a highly permissive license, allowing you to freely use the results from CHEESE Electrostatics for any purpose, including commercial use, at no cost.
Technical Insights: Our CHEESE Electrostatics technology drastically reduces the time required to compute electrostatic charges—what traditionally took 30 minutes to 10 hours now takes just a fraction of a second, maintaining up to a 0.98 correlation with DFT.
Understanding ESP and RESP Charges: Electrostatic Potential (ESP) charges are crucial for understanding molecular interactions as they are calculated to match the molecular electrostatic potential. RESP (Restrained Electrostatic Potential) charges add constraints to better mimic observed behaviors in molecules. Traditionally, computing these charges via DFT could take up to several hours for larger molecules. Now, with CHEESE Electrostatics, you can achieve these results in just a blink—without compromising on accuracy!
