Entropy and Selectivity
Most scientists would understand entropy as a thermodynamic property. However, applying “information entropy” to kinase-inhibitor profiling is a new way to solve the old problem of making sense out of the large amounts of data, according to Joost Uitdehaag, Ph.D., senior research scientist at Merck.
“Today, researchers are able to identify hits from new libraries of compounds using high-throughput screening. There is a lot of debate as to how to improve selectivity of these hits in the process and to determine when your compound is sufficiently selective. But, it all starts with quantification, with being able to compare actual values. This is what the entropy score for selectivity brings to the field.”
Dr. Uitdehaag said this process allows one to very quickly choose the best compound to take forward into further testing. “We have proposed a way to calculate a single value from a set of IC50 data to quantify selectivity profiling from panel profiling. It is a powerful way to study molecular mechanisms of kinase inhibitor selectivity.
“Often, other methods are utilized such as dotting a kinome tree, heat maps or a radius plot, but these only provide a qualitative comparison. For quantitative approaches, others have developed a selectivity score based on kinase-profiling data. But this doesn’t provide sufficient sensitivity. Other common methods include using the so-called Gini score or a partition index. None of these measures are fully adequate.”
How do these new equations work? They are based on the principle that an inhibitor candidate will assume a Boltzmann distribution across the various targets when confronted by multiple kinases. “This distribution has calculated entropy. If it is, for example, 2.2, which is an average measure of selectivity, the compound has average selectivity. If the calculation ends up to be 1, however, this indicates that the compound is a much more selective inhibitor. One can use this information to quickly select the best candidates after screening. The nice thing is that this method gives consistent values across profiling experiments, so it’s really general.”
Dr. Uitdehaag also noted that selectivity entropy can be used to study the success of candidates in clinical trials. “We assessed clinically tested inhibitors and determined their selectivity scores. We found that the most successful compounds actually are those with more broad selective profiles. These findings indicate that selective candidates have less of a chance for surviving early clinical trials.”
Indicating the selectivity of an inhibitor should ultimately be just as commonplace as indicating its IC50, advised Dr. Uitdehaag. “I think, for instance, the selectivity values of inhibitors should be reported when people do biological validation experiments with them. It would make a lot of sense.”
Although great advances have been made over the last decade in measuring and predicting kinase inhibitor selectivity, a number of issues remain. Advances in structural-guided modeling and in enhancing the selectivity of assays should provide critical improvements for future drug development and therapeutic target expansion.