Fuzzy Context Specific Matched Molecular Pairs Vincent Le Guilloux & Peter Schmidtke [email protected] [email protected] Discngine SAS, Paris, France 01 Introduction Matched Molecular Pairs (MMP) analysis [1 ] is frequently used to study the impact of chemical transformations on a given outcome. Identifying MMPs requires a shared structural part on two distinct molecules. Large amounts of 02 Classical MMP experimental data is necessary to derive statistically meaningful context specific rules [2]. Here the use of reduced graphs [3] is presented to augment statistical significance of extracted transformation rules. vs Fuzzy Context Specific MMP I .. Molecular Representation Molecular Structure Reduced Graph II .. Recursive Fragmentation >> Solubility 1 00 % >> Activity vs Target 90 % 10 % Solubility 1 00 % >> III .. MMP Enumeration Activity vs Target Pharmacophore Legend IV .. MMP Analysis Solubility Acceptor & Donor Activity vs Target 1 00 % 60 % 30 % 10 % Aliphatic 60 % 30 % 10 % - Aromatic Ring Acceptor Donor Neg. Charged + Pos. Charged 03 Use case: Kinase compound optimization Here an example optimizing GSK-3b compounds is shown considering LCK as monitored off-target. Molecular solubility was also considered. Using Chembl 1 6 [4], a total of 270 compounds have been selected with experimentally measured activities versus GSK-3b and LCK. MMPs have been calculated using classical MMP and fcsMMP with a context specificity up to four bonds. Statistics for replacement rules are shown and an example of a fcsMMP graph query is shown using Discngine Network. 04 Methods Subset of fragment replacement rules for fcsMMPs with a context up to four reduced graph bonds. 39 46 Context 27 52 24 22 Number of chemical transformations observed more than 20 times in the data. Frag. 1 Frag. 2 GSK-3b Solubility LCK 01 n >> 24 >> 1 0 >> 1 9 >> 1 2 >> 1 7 >> 1 2 Generate reduced graph of input molecule from chemist 03 02 04 Fragment the reduced graph Query & Vizualization of fcsMMP graph 05 23 11 >> 1 2 >> 1 2 >> 1 0 >> 8 ... 3 06 Fragment replacement rules for classical MMP analysis. Number of chemical transformations observed more or equal than 3 times in the data. Molecules and activity datasets were retrieved from Chembl 1 6[X]. Generation of reduced graphs [2] of molecules, detection of MMPs was done with the Discngine Chemistry Collection for Pipeline Pilot. 1 2 >> Generation and vizualization of the fcsMMP graph was performed using the Discngine Network Collection for Pipeline Pilot. Solubility was calculated using the Chemistry Collection of Pipeline Pilot. Retrieval of molecular structures and activity ranges for all matches found in the graph. 05 References [1 ] [2] [3] [4] Dossetter et al. DDT, 1 8, 201 3 Papadatos et al. JCIM, 50, 201 0 Birchall et al. Methods Mol Biol, 672, 201 1 Gaulton et al. NAR, 40, 201 1