Opportunities with large-market diseases are rapidly shrinking and experimental spaces are rapidly growing, so how can you even maintain, let alone accelerate, profitable drug discovery? Quantitative Medicine’s Computational Research Engine™ is a pivotal part of the solution. The CoRE™ can help do what it takes to succeed in today’s drug discovery field:
Innovate. Success in today’s pharmaceutical industry requires innovation. It requires going beyond predictive modeling to active machine learning, the basis of the CoRE™’s unique abilities.
Accelerate. Increasing research efficiency is vital, not only to maintain a profitable drug discovery campaign, but to accelerate it. A more complete characterization of the compounds’ effects against multiple, diverse targets accomplished with exponentially fewer experiments yields better results at a more affordable cost.
Integrate. The CoRE™ can use data from different experimental methods, across very different phases in the drug discovery and development process and can therefore operate within the existing pipeline structure of drug discovery.
Collaborate. Interaction from all sides fosters improvement to the drug discovery and development process as a whole and supports continuity throughout. The CoRE™ actively facilitates collaboration spanning not only divisions across an organization, but pre-competitively across an entire industry.
Going beyond predictive modeling to active machine learning
The use of computational predictive methods in the process of drug discovery and development is called virtual screening. A virtual screening process often starts by building a model from a small selection of experiments – the “training set” – to predict the results of unobserved experiments. While building these models, structural information like quantitative structural activity relationship (QSAR) data is sometimes used. After these models are built, a much smaller number of experiments are executed to confirm predictions made by the model.
Errors made by these predictive models can be costly. False positive results lead to wasted experimental effort characterizing ineffective compounds, while false negatives are potentially missed opportunities.
In order to address these issues, there have been significant efforts made to improve the predictions, by improving the algorithms used to create the predictive models. A predictive model can only be as good as the data used to train it, therefore solely improving the algorithms has limits as well. Additional efforts have been made to improve the throughput of experimental methods, enabling more data to be generated with less expense. The increase in available data has yielded improvements in the model accuracy, but researchers continue to be disappointed by diminishing returns as the amount of data has increased. Active machine learning can dramatically improve this situation.
Active machine learning is an exciting advancement in machine learning, employing an iterative, predictive model-building cycle. The process can direct experimentation, to uncover the most important, missing information, so that the accuracy of these models can be rapidly improved. Adopting active machine learning results in far less experimentation being needed to yield very accurate results.
Increasing efficiency to accelerate drug discovery and reduce costs
Quantitative Medicine provides a solid solution to this problem. An advanced, big data analytics technology that utilizes cutting-edge active machine learning methods to dramatically reduce the time and cost of the drug discovery and development process. Applicable in the initial stages of the process, the CoRE™ was not designed to replace experimentation with computation, but to complement current industry practices by using computation to guide experimentation.
The Computational Research Engine™ can analyze a large number of drug discovery factors concurrently. For example, both desired effects and side effects, such as toxicity, could be considered simultaneously. The CoRE™ uses many diverse sources of experimental data and a proprietary combination of machine learning methods to develop predictive models that effectively direct the discovery and development of compounds with the greatest potential for success.
This innovative approach allows the Computational Research Engine™ to efficiently determine which compound characteristics produce the right effects, and subsequently recommend how to optimize compounds for those effects. With this capability, those compounds with increased function and fewer side effects can be rapidly identified. In one study, the model was accurate enough to direct experimentation to find 55% of the hits in an experimental space after sampling only 3% of the data. Learn more about how CoRE™ can further your team’s success in the drug discovery process.