Computational models for life science range from empirical data mining relationships among key variables to mathematical simulations of known physical processes. Ultimately, the value of computational models can be assessed simply: can they be used for actual prediction of relevant biological activity of new chemical entitities that exist only in a computer? One very simple, and surprisingly useful, formulation of the question is the idea of neighborhood behavior: if two samples are similar enough in terms of computational model values, are they highly likely to also be similar in actual biological activity?

Similarity studies are often essential as part of turning the data deluge into new knowledge. Which measured, or computed, variables are in fact important? Can we leap from one chemical series, or from one phenotypical expression pattern, to another with any confidence?

These abilities are crucial in chemical structure design, both de novo and for searching databases of structures with measured activity. They are also crucial for deconvolution of gene and protein expression patterns.

The projects undertaken at Vistamont Consultancy focus on defining the crucial elements of similarity, including:

• computational molecular descriptors such as shape, pharmacophores, hydrophobicity
• for QSAR prediction
• for Safe Exchange of Chemical Information
• for database searches that suggest potential activity
• computational biological descriptors
• for similarity-based estimation of clinical application
• for deconvolution and estimation of molecular targets in pathways

Over the past two decades, as Senior Fellow with a major cheminformatics and chemical synthesis company, Tripos, the founder and Sole Proprietor of Vistamont Consultancy helped chemical representations advance from 2-D structural cartoons into 3-D descriptors of shape and charge and to 3-D pharmacophoric descriptions. He was one of the creators of CoMFA(tm), the major 3-D QSAR method in cheminformatics, and played an early role in focusing the idea of chemical diversity toward its useful application to biological projects via chemical libray design.

Since leaving to form Vistamont Consultancy, David Patterson has strengthened the data mining connections to drug discovery and has extended his patented definitions of meaningful similarity into the regime of proteomics and genomics.

Vistamont Consultancy, calling as necessary on colleagues in the fields of computational chemistry, data mining and biotech, offers consulting services that range from short individual studies to long term collaborations.