Network Collection 2016 brings Formal Concept Analysis to your desktops / by Peter Schmidtke

We are pleased to announce the release of the Network Component Collection 2016 for Pipeline Pilot. Among a few improvements we are very excited to bring our first components unleashing the power of formal concept analysis (FCA) on your desktops & servers. Take also a look on our new features in the Chemistry Collection.

What can you do with formal concept analysis (FCA)?

This analytical method allows to gain deeper insights into your data by enabling currently underutilized datamining paradigms in cheminformatics. The scope of FCA is very large, but we have focused on bringing FCA to cheminformatics and screening data analysis as this is what we know best! 
Nevertheless, you can use our generic FCA components to perform analyses in various other fields where data-mining is important.

The brand new release comes with components and protocols allowing natural fuzzy clustering...give it a try, your Medicinal Chemists will love it. Furthermore we are very soon releasing the app built in collaboration with GSK to perform target deconvolution of phenotypic screening results. Results from this collaboration have already been presented by Nicola Richmond (GSK) at the German Conference on Cheminformatics (GCC2015) in Fulda, Germany.

Example of clustering result using formal concept analysis and the Discngine Network Collection.

What is formal concept analysis (FCA)?

FCA originated in the early 80's in the group of Wille, Ganter & Burmeister at the university of Darmstadt. The principle aim is to organize data composed of objects and their attributes into a concept hierarchy. A concept is a set of objects that all share the same set of attributes and as such, FCA is an excellent way for performing human understandable and natural multi-parametric clustering tasks. The output of FCA is a hierarchically ordered collection of formal concepts, a formal concept lattice. This lattice also allows to derive very useful attribute implications. 

What do you get with this release?

Within the 2016 release you will get access to three new components allowing to build:

  • Generic concept lattices between whatever sets of user-defined objects and attributes (strings)
  • Molecular concept lattices, where objects are molecules and attributes are fragments of the molecules following user-defined fragmentation schemes
  • Phenotypic screening lattices, which is a component integrating a few nifty bits and pieces using FCA for performing target deconvolution of phenotypic screening assays

The collection also includes sample protocols showing how to use the components. Two sample applications will be released shortly and will include functional apps showing compound clustering and phenotypic screening results analysis.

Compatibility

This release is officially supported for Pipeline Pilot 2016 & 9.x.