Other Languages

PIPELINE PILOT

Empower your research team with a flexible scientific platform that drives efficiency, collaboration and innovation.

The player will show in this div

Advanced Modeling Collection for Pipeline Pilot

The Pipeline Pilot Advanced Modeling component collection provides methods for Recursive Partitioning (RP) and Multi-Objective Pareto Optimization as well as Genetic Function Approximation (GFA) QSAR analysis.

  • The Recursive Partitioning components provide a variety of RP methods including single tree and forest of trees learners. The methods can learn on single or multiple response variables.
  • The Pareto Optimization components provide methods for multi-objective optimization problems and provide solutions whose criteria trade off amongst two or more partially conflicting goals.
  • Genetic Function Approximation (GFA) applies a sophisticated genetic algorithm method to calculate QSARs. These identify critical relationships between properties and the characteristics in a set of molecules.

With the Recursive Partitioning Components you can:

  • Perform very rapid learning and data mining experiments on very large data sets with very large numbers of descriptors
  • Learn molecular data sets using fingerprints as descriptors
  • Visualize trees to understand the relationships between descriptors and responses
  • Analyze descriptor usage to identify the most discriminating descriptors
  • Rapidly apply models to predict new data sets

With the Pareto Optimization Components you can:

  • Optimize solutions for problems as diverse as combinatorial library design, formulation ingredient optimization or stock portfolio risk management
  • Find individual samples with a data set that have the best trade-off of desired property values
  • Find subsets of samples from a larger data set that collectively have the best trade-offs between desired property values

With GFA Analysis you can:

  • Return multiple models rather than a single "best" model by creating a great number of trial models
  • Increase the likelihood of encountering spurious correlations specific to the training data that make a model appear better than it is in reality
  • Leverage this technique to generate hypotheses rather than test hypotheses

Next Steps

Contact Us

Browse By:

Learn More

Resource Center

Visit our Resource Center
Stay connected with us:
US: +1 888 249 2284
|
Europe: +44 1223 2285000
|
Asia Pacific: +81 3 3578 3860
Smarter Science. Better Business.