Data Driven Drug Discovery & Development - a computational biology perspective
The pharmaceutical industry is challenged by high cost and attrition rates. To break this cycle we must improve our ability to use available information. I will illustrate this by taking a climb to the top from early drug discovery through development. Examples from computational biology will be used along the journey: D1 – phenotypic screening is the primary source for first in class NMEs but poses the challenge of deconvolution. Historical bioactivity data can be used to help solve this puzzle. D2 – high throughput screening is the workhorse for hit identification but poor on content. A workflow, cherry picking positive wells for high content imaging together with automated image analysis, allows identification of artifacts and improves the ability to make decisions in HTS. D3 – the DDR pathway provides attractive oncology targets. A mechanistic cell cycle model can explain in vitro behavior and in vivo observations of combination treatment to inform clinical design. D4 – safety concerns were raised when a clinical study noticed drop in neutrophil counts. A multiscale model was developed and led to the conclusion that neutrophil blood count is expected to saturate at high dosages. D5 – gamma secretase inhibitors for Alzheimer’s disease have been observed not to work – or even make patients worse. This conundrum can be explained by a tale of three secretases. This boosts the confidence in current AZ programs. Together these examples show why the quantitative use of data is a must in drug discovery and development.