Structure generation, metabolite space, and metabolite likeness
1 like510 views
The document discusses metabolite identification from mass spectrometry data. It describes how a structure generator works to generate candidate metabolite structures from an elemental composition. The generator adds bonds in all possible ways to create structures, then uses isomorphism and canonical labeling to remove duplicate structures within the same isomorphic class. This process generates a list of candidate metabolite structures for further analysis and filtering against experimental data.
Structure generation, metabolite space, and metabolite likeness
- 1. Structure Generation, Metabolite Space, and Metabolite-Likeness Julio E. Peironcely @peyron Juliopeironcely.com PhD student at Leiden University and TNO
- 2. Metabolomics the quantitative and qualitative analysis of all metabolites in samples of cells, body fluids, tissues, etc. Julio E. Peironcely
- 3. Metabolomics Experi- Biological Biological Sample Data Data pre- Data mental Sampling inter- question preparation acquisition processing analysis design pretation Metabolites Relevant biomolecules/ List of Samples Raw data connectivities Protocol peaks/ & biomolecules Models Julio E. Peironcely
- 4. Metabolomics Experi- Biological Biological Sample Data Data pre- Data mental Sampling inter- question preparation acquisition processing analysis design pretation Metabolites Relevant biomolecules/ List of Samples Raw data connectivities Protocol peaks/ & biomolecules Models Julio E. Peironcely
- 6. We have Elemental Composition Fragments (sometimes) Experimental Information Julio E. Peironcely
- 7. We want List Of Candidate Structures As Short As Possible Good Structure Is In The List Julio E. Peironcely
- 8. We need Structure Generator Keep only metabolites Use experimental information to filter molecules Julio E. Peironcely
- 9. Elemental Composition Julio E. Peironcely
- 10. Elemental Composition Structure Generation Julio E. Peironcely
- 11. Elemental Composition Structure Generation Molecules Julio E. Peironcely
- 12. Structure Generator Elemental Fragments Formula Generate Candidate Structures In collaboration with Jean-Loup Faulon, Evry University Julio E. Peironcely
- 13. Structure Generator Elemental Fragments Formula Generate Keep Molecules if Canonical Augmenta:on Candidate Structures In collaboration with Jean-Loup Faulon, Evry University Julio E. Peironcely
- 14. Structure Generator Adding bonds In collaboration with Jean-Loup Faulon, Evry University Julio E. Peironcely
- 15. Structure Generator Isomorphism Isomorphic class Isomorphic class triangle + 1 edge 3-edge chain 1 1 1 1 1 2 3 2 3 1 2 3 2 3 1 3 1 2 3 4 2 4 4 4 1 3 2 3 4 2 4 3 2 3 4 2 4 4 4 In collaboration with Jean-Loup Faulon, Evry University Julio E. Peironcely
- 16. Structure Generator Isomorphism Isomorphic class Isomorphic class triangle + 1 edge 3-edge chain 1 1 1 1 1 2 3 2 3 1 2 3 2 3 1 3 1 2 3 4 2 4 4 4 1 3 2 3 4 2 4 3 2 3 4 2 4 4 4 Output ONLY orange graphs In collaboration with Jean-Loup Faulon, Evry University Julio E. Peironcely
- 17. Structure Generator Canonical Labeling 1 1 1 1 1 2 3 2 3 1 2 3 2 3 1 3 1 2 3 4 2 4 4 4 1 3 2 3 4 2 4 3 2 3 4 2 4 4 4 Canonizer (Nauty) (1,2) (1,3) (1,4) (1,2) (1,3) (2,4) (2,3) Julio E. Peironcely
- 18. Only 1 canonical labeling in each isomorphic class
- 19. Use canonizer to 1 remove duplicates after 2 3 5 4 each extension 1 (1,2) 2 3 4 5 1 1 (1,2)(1,3) 2 3 2 3 4 5 4 5 (1,2)(1,3)(1,4) (1,2)(1,3) (2,3) 1 1 1 1 2 2 3 2 3 2 3 1 3 5 5 4 5 5 4 4 4 2 3 (1,2)(1,3)(1,4) (1,2)(1,3)(1,4) (1,2)(1,3)(1,4) (2,3) 4 5 (3,4) (4,5) (1,2)(1,3) X (2,3)(2,4)
- 20. Canonical Augmentation A canonical object augmented in a canonical way produces a canonical object Julio E. Peironcely
- 21. Check For Canonical Augmentation Keep object if a canonical deletion takes you to the canonical father Julio E. Peironcely
- 22. Accept only canonically 1 2 3 augmented graphs 4 (1,2) 5 1 2 3 4 5 1 1 (1,2)(1,3) 2 3 2 3 4 5 4 5 (1,2)(1,3)(1,4) (1,2)(1,3) (2,3) 1 1 1 1 2 2 3 2 3 2 3 1 3 5 5 4 5 5 4 4 4 2 3 (1,2)(1,3)(1,4) (1,2)(1,3)(1,4) (1,2)(1,3)(1,4) (2,3) 4 5 (3,4) (4,5) (1,2)(1,3) X (2,3)(2,4) X
- 23. Structure Generator Results MOLGEN same # of molecules p-Cresol Glycine Phenylalanine Malic acid D-Cysteine sulfate Elemental C2H5NO2 C9H11NO2 C4H6O5 C3H7NO2S C7H8O3S Composition # Output 84 277,810,163 8,070 3,838 10,203,389 Molecules 6 4,037,499 1,601 100 19,940 1 Fragment 93,137 948 2 Fragments 584 3 Fragments 278 In collaboration with Jean-Loup Faulon, Evry University Julio E. Peironcely
- 24. Lots of candidates structures
- 25. We are looking for metabolites
- 26. Elemental Composition Structure Metabolite Generation Likeness Molecules Julio E. Peironcely
- 27. Elemental Composition Metabolites Structure Metabolite Generation Likeness Molecules Julio E. Peironcely
- 28. How do metabolites look like? Understanding and Classifying Metabolite Space and Metabolite-Likeness Julio E. Peironcely et al. PLoS One (in press)
- 29. HMDB ZINC 8K 21M Julio E. Peironcely
- 30. metabolites non metabolites Water Solubility MW C Atoms Struc. Complexity PSA Julio E. Peironcely
- 31. PCA Julio E. Peironcely
- 32. PCA
- 33. Not so different
- 34. Decision Tree Julio E. Peironcely
- 35. Elemental Composition Metabolites Structure Metabolite Generation Likeness Molecules Julio E. Peironcely
- 36. Metabolite-likeness Representation + Classification HMDB ZINC 8K 21M Atom Counts Physicochemical desc. Support Vector Machines (SVM) MDL Public Keys Random Forest (RF) FCFP_4 Na誰ve Bayes (NB) ECFP_4 Julio E. Peironcely
- 37. Metabolite-likeness HMDB 8K ZINC 21M Standardization Atom Counts Diversity Selection Physicochemical desc. MDL Public Keys FCFP_4 ECFP_4 Julio E. Peironcely
- 38. Metabolite-likeness HMDB 8K ZINC 21M Standardization Atom Counts Diversity Selection Physicochemical desc. MDL Public Keys FCFP_4 Training Set Test Set ECFP_4 532 + 532 6.4K + 6.4K Julio E. Peironcely
- 39. Metabolite-likeness HMDB 8K ZINC 21M Standardization Atom Counts Diversity Selection Physicochemical desc. MDL Public Keys FCFP_4 Training Set Test Set ECFP_4 532 + 532 6.4K + 6.4K 5-fold CV SVM RF BC Julio E. Peironcely
- 40. Metabolite-likeness HMDB 8K ZINC 21M Standardization Diversity Selection 3 classifiers X Training Set Test Set 5 descriptions 532 + 532 6.4K + 6.4K 5-fold CV Metabolite likeness SVM RF BC Julio E. Peironcely
- 41. Metabolite-likeness HMDB 8K ZINC 21M Best = RF MDLPublicKeys Standardization Sensitivity Specificity AUC Diversity Selection 99.84% 87.52% 99.20% Training Set Test Set Bad BC P_desc 532 + 532 6.4K + 6.4K Sensitivity Specificity AUC 5-fold CV Metabolite likeness SVM RF BC 42.51% 86.56% 61.57% Julio E. Peironcely
- 42. Metabolite-likeness, external validation HMDB External DrugBank ChEMBL validation set Random Selection Standardization Metabolite likeness Julio E. Peironcely
- 43. Metabolite-likeness, external validation Julio E. Peironcely
- 45. Met-likeness + structure generation (malic acid) 8K 100% 57% 77% Julio E. Peironcely
- 46. Met-likeness + structure generation (methylhistamine) 260K 71% 46% Julio E. Peironcely
- 47. What else do we know about our molecules?
- 48. Molecule Minimized_Energy ALogP Index Phenylalanine 0.1100 -1.605 5142
- 49. Molecule Minimized_Energy ALogP Index C9H11NO2 0.1100 -1.605 5142 Structure Generation 277 M Julio E. Peironcely
- 50. Molecule Minimized_Energy ALogP Index C9H11NO2 0.1100 -1.605 5142 99% Structure Generation 44% 41 K Julio E. Peironcely
- 51. Molecule Minimized_Energy ALogP Index C9H11NO2 E < 10 0.1100 -1.605 5142 Structure Generation 40% 8K Julio E. Peironcely
- 52. Molecule Minimized_Energy ALogP Index C9H11NO2 E < 10 0.1100 -1.605 5142 ALogP < -1 Structure Generation 76% 31 Julio E. Peironcely
- 53. Conclusions Met-Likeness prediction is good, interpretation not Local models needed Structure Generator + Met-Likeness + other constraints = Met Id improvement Julio E. Peironcely
- 54. Acknowledgements Leiden University University of Cambridge Miguel Rojas-Cherto Andreas Bender Piotr Kasper Michael van Vliet Theo Reijmers Rob Vreeken Evry University Ronnie van Doorn Jean-Loup Faulon Thomas Hankemeier Davide Fichera TNO Quality of Life Leon Coulier Albert Tas HMP University of Alberta David Wishart Ying (Edison) Dong Julio E. Peironcely