Matched Molecular Pairs (MMP) Nodes updated

Community Extensions
1

The long-promised update to the MMP nodes has been released to the nightly build this morning.  We would strongly encourage anyone using the existing versions of the nodes to convert workflows to the new version, and also let us know if you encounter any issues.  We will release to the stable builds in around 1 week.

This is a major overhaul of the MMP Nodes which has been described at recent Cheminformatics SIGs and the Berlin Spring 2017 UGM.  Many of the original nodes have been deprecated, a number of new nodes and features added, and many of the new nodes are now streamable.

  • All Nodes
    • Performance improvements
    •  New predefined fragmentation patterns ('MATSY', 'Peptide Sidechains', 'Nucleic Acid Sidechains')
    • Improved rSMARTS validation
      • Now requires the bond between the two atoms to be both single and acyclic
      • Allows more complex SMARTS atoms matches (including recursive SMARTS)
      • No longer requires the SMARTS to be provided as an rSMARTS - although this is still supported for backwards compatibility
      • No longer requires atoms to be numbers in the SMARTS match 
  • Filtering (Updated)
    • Updated faster implementation
    • Streamable
    • Previous versions deprecated
  • Fragmentation nodes (UPDATED)
    • Performance improvements
    • Memory leaks fixed
    • AP Fingerprint columns have properties describing creation settings to avoid mismatch
    • Graph Distance AP fingerprint vectos added
    • ID can now be the RowID
    • HiLite mapping optionally implemented
    • New Progress View
    • Pass-through of data columns
    • Rendering of fragmentation
    • Complexity filter
    • Behavioural change to remove explicit H's for 1 cut option
    • Additional option regarding incoming explicit H's
    • Streamable
    • '3rd Gen' version deprecated
  • Pair Generation (UPDATED & NEW)
    • Performance improvements including parallelisation
    • HiLite mapping optionally implemented
    • Transforms can be filtered by
      • Heavy-Atom-Count change
      • Graph distance similarity
    • Remove explicit H's from pair output no longer available as an option
    • Added option to require attachment points to be attached by single, acyclic bonds in SMARTS pattern
    • Pass-through, difference (L-R or R-L) and ratio (L/R or R/L) of data columns from left or right of transform pair
    • New 'Reference Table' version of pair generation node implemented, only returning transforms between the rows of the two input tables
    • Deprecated older version
  • Apply Transforms (NEW)
    • New node to apply table of rSMARTS transforms
    • Allows AP-fingerprint environment filtering
    • Node Views showing progress, including currently processed transform(s)
    • Limited stereochemistry generation handling at present time
  • Rendering Nodes (NEW)
    • Show matching bonds (also calculates number of matching bonds)
    • Show cuttable bonds (also calculates number of cuttable bonds)
    • Streamable

Steve

2

Hi, many thanks for the updates, but could you show a a working example of how to use them best?

Much appreciated/Evert

3

Evert,

Certainly!  I will get a couple of example workflows set up and post back!

Steve

4

Hi Steve,

On your slides from the Knime UGM in Berlin you have a number of very interesting MMAnalyser workflows:

https://www.knime.org/files/spring-summit2017/Vernalis%20KNIME%20Spring%20summit%202017.pdf

Are these proprietary are or could you share these with the community?

Many thanks/Evert

5

So, we have made a number of minor bug fixes, and are now on v1.11.2 (Nightly build) and v1.11.1 (Stable builds) - all should retain full backwards compatibility with the original v1.11.0 release.  v1.11.2 should become available on the stable builds (v3.2-v3.4) later today.

Regards the Berlin workflows - I'm afraid those are proprietary, although there is little that cannot be derived from the peer-reviewed literature in the area.  They also used the older versions of the nodes and achieved the Matched Molecular Series (MMS) pieces with a lot of grouping/manipulation, and so arent really suitable for release anyway.  We now have internal nodes to do much of the MMS piece directly - those will likely be released once they have some remaining wrinkles ironed out!

As for examples, I realised that one of the two workflow relies on some new fingerprint nodes which we have had release permission for for some time, but not yet done. in order to show full functionality.  There is probably a couple of days of final sorting of those nodes, and so I will release the examples once those are also released

Steve

6

This topic was automatically closed 90 days after the last reply. New replies are no longer allowed.