Recent Advances in Computer-Aided Interpretation of 2D Representations of 3D Molecules

Transcription

1 ecent Advances in Computer-Aided Interpretation of 2D epresentations of 3D Molecules Karl edwed, Gregory M. Banik, Ph.D., Ty Abshear, Keith Kunitsky & Michelle D ouza, Ph.D. Bio-ad Laboratories, Inc. Informatics Division August 16, 2017 InChI I, Bethesda, MD

2 Many ources, Many Chemists, Many tructures

3 ption (A) for Merging tructures: tandardization riginal After

4 ption (B) for Merging tructures: Interpretation riginal There Is o After

5 Wikipedia Page for Pyranose Fischer Projections aworth Projection Chair epresentation ource of Image: - Accessed July 10, 2017.

6 aworth Projection

7 Chair

8 isyphean Task: Changing Chemists Drawing abits tructure drawing conventions such as boat/chair cyclohexane, Fischer projections, aworth projections, and 2.5D representations are deeply ingrained in all chemists today Every newly minted chemist in the world is being trained to draw structures with these conventions Changing the drawing habits of all current and future chemists is the definition of a isyphean task ource of Image: - Accessed July 16, 2017.

9 Computer Interpretation of 2D tructures The Problem: eretofore, most chemical software packages have been unable to accurately interpret traditional 2D representations of 3D molecules ecent Advance: Bio-ad s KnowItAll software has been reengineered to accurately interpret traditional 2D representations of 3D molecules by adding the ability to understand the sometimes subtle 3D intentions of 2D drawing conventions including: Interposition Foreshortening Perspective caling Ambiguous stereocenters are detected

10 KnowItAll ChemWindow CIP tereodescriptors Cahn-Ingold-Prelog Assignments / E/Z M/P tereochemical Depictions ecognized Boat / Chair epresentations Fischer Projections 2.5D (Pseudo 3D) Projections aworth Projections e-chemist s ash/wedge epresentations elical stereochemistry tereogenic planes (e.g., cyclophanes) Br Br Br Br tereogenic axes (e.g., sterically-hindered o-substituted biphenyls) Z PPh 2 M PPh 2 P E P PPh 2 PPh 2 M

11 PubChem ubstances Considered for tudy 200,000,000 Ending PubChem ubstance ID analyzed 92,207,320 Unused PubChem ubstance IDs 107,792,680 Used PubChem ubstance IDs 665,837 PubChem ubstances with no structure 15,559,322 PubChem ubstances with connection tables but no coordinates 27,781,844 PubChem ubstances with structures already normalized 194,249 PubChem ubstances that failed to generate an InChI 63,591,428 PubChem ubstances with valid, un-normalized structures 57,556,031 PubChem ubstances without defined stereocenters 6,035,397 PubChem ubstances with defined stereocenters

12 PubChem ubstances Considered for tudy 662,971 PubChem ubstances with 2.5D perspective drawings (10.98%) 51,038 PubChem ubstances with chair/boat rings (0.85%) 27,639 PubChem ubstances with aworth projections (0.46%) 2,015 PubChem ubstances with Fischer projections (0.03%) 6,035,397 PubChem ubstances with defined stereocenters

13 Interpretation Technology Validation ets Using a maximum diversity algorithm, 150 matched pairs of PubChem ubstances and PubChem Compounds were selected for each of the following categories: aworth Projections Fischer Projections Chair/Boat epresentations 2.5D epresentations Because the selections were diverse, they also tended to include some difficult examples

14 Validation Protocol Compare InChIKeys for both representations Investigate discrepancies PubChem ubstance PubChem Compound (tandardized) UYPITADKUEIY-ZDCXTMVA- Chemist manually adds stereobonds to standardized PubChem Compound to match the PubChem ubstance and generate InChI and InChIKey UYPITADKUEIY-ZDCXTMVA- KnowItAll perceives implicit stereochemistry from 2D representation and automatically adds stereobonds internally to original PubChem ubstance to generate InChI and InChIKey

15 elated Topic: CIP tereodescriptors Match,,,,,,

16 Validation Example Cyclohexane Chair Br Br Br E E Br PubChem ubstance (riginal) PubChem Compound (tandardized) PubChem Compound (Edited),,?,?,?,, PubChem ubstance (Interpreted),,

17 Validation Example aworth Projection with Misinformation 2 2 PubChem ubstance,,, PubChem Compound (tandardized),?,, PubChem Compound (Corrected),,,

18 Validation esults umber of records: 595 umber of stereocenters: 1,573 umber of matching stereocenters overall: 1,522 (96.76%) umber of matching / assignments: 1,512 (98.63%) umber of missed stereocenters: 20 (1.27%) umber of added stereocenters: 10 (0.64%) umber of stereocenters with reversed / assignments: 21 (1.34%)

19 Examples: Fischer Projections Te Te P

20 Examples: aworth Projections F s

21 Examples: Chair Projections P P F s s r r Br

22 Examples: 2.5D Projections C C r s s r r

23 ummary Part art form and part language, 2D representations of chemical structures have a very long history. Traditional structure drawing styles like cyclohexane chairs and boats, Fischer projections, aworth projections, and 2.5D perspective projections can be accurately interpreted by computer software, so e-chemists can end their battle against 127 years of scientific tradition, embrace the past, and live in peace and harmony with the rest of the chemical community.