maintains a strong focus on methods development (see "Technologies").
A constant thread is the use of probabilistic methods utilizing
all prior information to in selecting compounds to purchase,
screen, or synthesize. Previous development has included novel
Learning Methods software for the prediction of small-molecule
bio-activities, Bayesian statistical methods for interpretation
of results of high-throughput screens, decision-tree methods for
characterizing and augmenting high-throughput screening
libraries, surface-based phylogenetic tree generation of active
sites in protein families to help achieve drug specificity to
and within protein families, and a data-mining approach to
augment energetic considerations in interpreting results of
virtual screens. A current and representative internal
research topic discussed further in "Technologies"
is the use of probability distribution methods to formulate
new, mathematically rigorous approaches to the problems of
docking, structure-based 3D peptide profiling, pharmacophore
search, and flexible small-molecule alignments. The method yield
ensembles of solutions rather than individual solutions, helping
to circumvent imprecision that plagues current approaches.
also actively pursuing the GHKL superfamily of proteins to
accelerate early stage discovery against targets within this
superfamily. The superfamily includes diverse drug targets with
application to oncology (HSP90) and to infectious disease (Gyrase
B and Histidine Kinases). See Internal Discovery.