Sprouting out new ideas from knowledge ground

Posted by
Ákos Tarcsay
on 2018-09-12

Sprouting out new ideas from knowledge ground

We present a platform supporting the design of novel compounds by collecting the right amount of data into a unified design space. Our platform, Design Hub (Marvin Live's successor), provides an integrated solution, and combines raw data sources, predictive models and digested information into a single application. Besides, it efficiently supports streamlined design and triage of the hypothetical compounds. Without this list being exhaustive, the integrated knowledge includes physicochemical descriptors, combined metrics like the CNS MPO score, basic modeling capabilities and matched molecular pair (MMP) analysis. Visualizing data pre-processed by matched molecular pair analysis in straightforward, and conveniently transformed form has high impact on rational design. In this study we present a solution how MMP analysis can be applied to extend chemical intuition with computer-aided approach. This hypothetic study shows the potential evolution of a compound idea from the reference compound to a synthesis candidate through an example discovery project on hERG dataset. Data was extracted from ChEMBL [1] to mine chemical and related data and to generate matched molecular pairs. Fragmentation and matched pair generation were performed using MMPDB [2], whereas post-processing and visualization were implemented within the framework of Design Hub. The “smart assistant” capability presented here is nourished by available knowledge and is designed to improve decision-making on what kind of compounds to design or synthesize next.

[1] https://www.ebi.ac.uk/chembl/

[2] https://github.com/rdkit/mmpdb

Download the poster

 

We present a platform supporting the design of novel compounds by collecting the right amount of data into a unified design space. Our platform, Design Hub (Marvin Live's successor), provides an integrated solution, and combines raw data sources, predictive models and digested information into a single application. Besides, it efficiently supports streamlined design and triage of the hypothetical compounds. Without this list being exhaustive, the integrated knowledge includes physicochemical descriptors, combined metrics like the CNS MPO score, basic modeling capabilities and matched molecular pair (MMP) analysis. Visualizing data pre-processed by matched molecular pair analysis in straightforward, and conveniently transformed form has high impact on rational design. In this study we present a solution how MMP analysis can be applied to extend chemical intuition with computer-aided approach. This hypothetic study shows the potential evolution of a compound idea from the reference compound to a synthesis candidate through an example discovery project on hERG dataset. Data was extracted from ChEMBL [1] to mine chemical and related data and to generate matched molecular pairs. Fragmentation and matched pair generation were performed using MMPDB [2], whereas post-processing and visualization were implemented within the framework of Design Hub. The “smart assistant” capability presented here is nourished by available knowledge and is designed to improve decision-making on what kind of compounds to design or synthesize next.

[1] https://www.ebi.ac.uk/chembl/

[2] https://github.com/rdkit/mmpdb

Download the poster