One of the best ways to find new drugs is to leverage knowledge of previous attempts, successful and otherwise.
We have developed a user friendly web-based software to access 16 machine learning algorithms: Deep learning 3 layers (classification), Adaboosted decision trees (classification & regression), Bernoulli Naïve Bayes (classification), Laplacian Bayes (classification) & BayesianRidge (regression), K-nearest neighbors (classification & regression), Random forest (classification & regression), Support vector classification (classification & regression), LogisticRegression (classification) & ElasticNet (regression), XGBoost (classification & regression). We use ECFP & FCFP descriptors with diameter 4/6/8 along with bit folding and leave out validation options. We provide several data visualization options. Assay Central machine learning models can be used to filter and score compounds prior to testing. The following images illustrate how the classification and regression model statistics can be displayed for BACE1 datasets.
Benefits
- Making data accessible to machine learning
- Data intensive visualization resulting from these many models
- Closing the loop between experimentalist and data repositories
- Graphical display of models – instant feedback
- Model applicability – multiple methods to assess with scores and graphics.
Access
- We can use Assay Central in fee for service work for you.
- We can provide an annual license for you to access this software on your own server.
- We provide maintenance and customization options.
Success stories
We have worked on collaborative projects with companies on a fee-for service basis:
- Worked with a major US consumer product company to collate public estrogen receptor data and score their chemicals.
- With a major pharmaceutical company to model their ADME data and propose compounds to synthesize.
- With a preclinical CRO to model their internal drug screening data make predictions for future testing as well as evaluate blood brain barrier permeation.
- Multiple collaborations on whole cell and target specific models – have identified novel inhibitors for academic collaborators.
- Model building and testing with academics with access to previously unpublished data
- Built and validated transporter models for different probes and shared models.
We can work with you to automate the curation of your in house data and build machine learning models that you need to generate novel insights.
If you are a VC or company we can use our models to perform independent due diligence to evaluate in-licensing opportunities or score company pipelines in your portfolio.
If you are a company we can develop next generation leads using our models, library enumeration and retrosynthesis tools.
We have developed extensive capabilities using recurrent neural networks to design therapeutics de novo.
We have expertise with using our models with molecules large and small such as macrolactones and proteolysis targeting chimeras (PROTACs).
We have capabilities to custom develop approaches to model molecule properties as well as spectra.