Abstract:

Living systems and their biomolecules are well understood by atomic modeling of their structural chemistry, which has led to a profound revolution in the digitalization of biological systems. These digitized systems are being catalogued in online databases, analyzed and modeled computationally primarily by inference of homology with the known experimental counterparts. Such digitalization of biology is likely to have an immediate and dramatic impact in the area of drug discovery and development. Atomic level predictive bioanalytics that integrate heterogeneous data sources, systems structural biology algorithms and bioinformatics approaches to identify multiscale biological relationships of compound proteome interactions foreshadows a new era of faster, safer, better and cheaper drug repurposing and discovery. The use of predictive bioanalytics at the molecular level has several advantages allowing for a mechanism- and hypothesis-free exploration of potential drug interactions and, furthermore, makes possible the discovery of more complex and nuanced drug-target interactions. This also opens the door for novel approaches, such as phage antimicrobials, immune system reprogramming and regenerative medicine. Predictive bioanalytical tools also make use of vast data sets of biomedical data, enhancing the repurposing of drugs already approved by the FDA for human use. The repurposing of FDA-approved drugs is particularly attractive, because it might enable researchers to minimize the size and cost of clinical studies for the new uses of such drugs. In combination with large data, so- called 'Big Data', studies of the 'off-label' use of such drugs in the general population could further lead to novel approaches to drug safety that are more rapid and cost efficient than existing drug discovery pipelines. The predictions of candidate drugs could also be tailored to specific individuals based on available information regarding their proteome (from single nucleotide polymorphism data obtainable from companies such 23andme, or even whole-genome sequencing), to minimize adverse effects and cost, as well as increasing efficacy.

We have implemented a modeling and "big data" integration pipeline that generates an interaction network between 'all' drugs (currently 3,733 human ingestible drugs) and 'all' proteins (currently 48,278) as a representative of the protein universe, using a hierarchical chem- and bio-informatic fragment-based dynamics protocol (~ 1 billion predicted interactions evaluated, considering multiple binding sites per protein). I will lead the discussion by introducing our efforts in this area and the success obtained that leverages the evolutionary basis of small molecule and protein interactions and the vast amounts of digitized biomolecular data to predict efficiently candidate drugs for more than 2000 indications and acts as a 'plug-in' to evaluate such drug candidates in the search for novel treatments. It also provides a path towards applying key aspects of the digital world that are so successful in information technology to the biomedicine, potentially breaking the infamous Eroom's Law (i.e. Moore's law backwards) of pharmacotherapeutics, where drug development becomes ever more expensive, ever more slowly developed and ever less effective, and finally placing the search for new drugs and treatments on a Moore's Law-like curve leading to ever cheaper, safer, ever more rapidly developed and ever more effective pharmacotherapeutics.

Click here www.stat.purdue.edu/~doerge/BIOINFORM.D/SPRING16/sem.html for a full schedule of BIOINFORMATICS SEMINARS, past and present.