The Senomyx Discovery Process



Senomyx Mission:


To discover new and improved flavor and fragrance molecules through integrating our enhanced understanding of the biology of taste and smell with high throughput molecular discovery technology.



High Throughput Sensory Discovery:



A rapid, iterative process based on a system of tightly integrated and automated just-in-time components of molecular discovery discipline.

  Knowledge-based Computational Library Design

  High Throughput Synthesis

  Analysis and Purification

  Assay Development

  High Throughput Screening



Informatics Foundation and infrastructure:


  Oracle Database (Senobase)

  Daylight Oracle Cartridge and toolkits

  Modular server-side software components (VB, Java)

  Web-based user interfaces

  Numerous behind the scenes middleware that sets the foundation for rapid application development.


Informatics Tools:


·        Structure/substructure searching

·        List generation

·        List management

·        Method writer


Computational molecular design:


·           Diversity space is limited with matrix synthesis

·           Goal is to expand diversity space through cherry picking

·           Selection from virtual library

·           2-D topology-based methods

·           Nearest neighbor approach

·           Genetic algorithm used for searching

·           Cherry picking hundreds for synthesis

·           Fully integrated into workflow


Virtual libraries:


·           Defined by a generic reaction (A+B+C => ABC)

·           Scope of viable reagents

·           Novelty and relevance

·           The sum of all virtual libraries constitutes the search space for target guided design


Genetic Algorithm:


1.  Start with a randomly generated seed population of n candidate reactions or from a pre-existing population


2.  Repeat the following steps until n offspring have been    created:

·  Select a pair of parent candidates from the current population

·  Apply a singe-point crossover operator in a chemically meaningful way; keep one new offspring randomly

·  Mutate the offspring with specified mutation probability


3.  Combine new and old populations


4.  Calculate the fitness of each candidate in the combined population


5.  Sort population by fitness and cull the less fit half.


6.  Go to step 2.




Fitness based selection:


•Maximize activity in biological assay

Use a “training set” of compounds whose activities have been measured in a biological assay.  Estimate the activity of each new compound by selecting its 10 nearest neighbors from the training set and taking a weighted average of their activities.  Nearest neighbors are determined by Tanimoto coefficients of similarity for 2D binary fingerprint comparison.

•Maximize population diversity

        Diversity of a given set of compounds D(S) is estimated as





         where n is the number of compounds in S and d is the diversity between two

         molecules based on Tanimoto coefficient for binary fingerprint comparison

        Ref. Dimitris K. Agrafiotis & Victor S. Lobanov, J. Chem. Inf. Comput. Sci. 1999, 39, 51-58

•Optimize reagent use




Automated library synthesis:


·        Chemistry definition  (reaction transform – SMIRKS – and building block lists)

·        Library enumeration (GA)

·        Run preparation (material balance)

·        Chemistry services (bulk to automation ready)

·        Automated synthesis

·        Plate reformatting

·        Purification and LC/MS analysis

·        High throughput screening





  Design, synthesis, and screening cycle is ~1 week

  Current throughput is several thousand compounds per week

  >50,000 compounds have been synthesized and screened using more than 40 automated combinatorial reactions

  Hundreds of active molecules have been identified from a total of over 10 chemical classes