Practical examples of the new metabolism
ranking system in Meteor Nexus
Dr. Ernest Murray
Senior Scientist
[email protected]
Outline
• Introduction to Meteor Nexus
• Overview of the new Ranking System
• Occurrence Ratio Method
• Metabolism database
• Query Specific Occurrence Ratio
• Example predictions
• Summary
Meteor Nexus
How Meteor Nexus Works
Knowledge base
Dictionary of
biotransformations
What reactions could occur?
Rule base
How likely that each
reaction will occur?
Processing constraint
Rule Base
• Biotransformation ranking is determined by a reasoningbased interpretation of two types of rules describing
Absolute likelihood of a
single biotransformation
Relative likelihood of a pair
of biotransformations
Probable
Plausible
Probable
Equivocal
Doubted
Improbable
WG Button et al, J Chem Inf Comput Sci 43 371–1377 (2003)
Probable
Meteor Nexus
Predicted metabolic tree
Meteor Nexus
Predicted metabolic tree
Filter options and
supporting evidence
Lasofoxifene: Meteor Nexus Prediction
O
O
In Meteor Nexus the biotransformation
cut off is set using the reasoning level.
N
BT Number
243
66
222
253
245
67
78
76
27
20
234
235
34
69
233
77
Name
Oxidative N-Dealkylation
Hydroxylation of Alicyclic Methylene Adjacent to Aromatic Ring
Oxidation of 4-Alkylphenols
Oxidative O-Dealkylation
Oxidative N-Dealkylation
Lactams from Aza-Alicyclic Compounds
Para Hydroxylations of Monosubstituted Benzene Compounds
Ortho Hydroxylations of Monosubstituted Benzene Compounds
Glucuronidation of Aromatic Alcohols
O-Sulphonation of Aromatic Alcohols
6-Hydroxylation of 1,2,4-Trisubstituted Benzenes
5-Hydroxylation of 1,2,4-Trisubstituted Benzenes
Glucuronidation of Tertiary Aliphatic Amines
Hydroxylation of Unfunctionalised Alicyclic Methylene
3-Hydroxylation of 1,2,4-Trisubstituted Benzenes
Meta Hydroxylation of Monosubstituted Benzene Compounds
Reasoning Level
PROBABLE
PLAUSIBLE
PLAUSIBLE
PLAUSIBLE
PLAUSIBLE
PLAUSIBLE
PLAUSIBLE
EQUIVOCAL
EQUIVOCAL
EQUIVOCAL
EQUIVOCAL
EQUIVOCAL
EQUIVOCAL
EQUIVOCAL
DOUBTED
DOUBTED
Meteor Nexus New Ranking System
Knowledge base
Dictionary of
biotransformations
Database
Experimental
reactions
Expert
system
What reactions could occur?
Machine
learning
How likely is it that each
reaction will occur?
Processing constraint
Metabolism Database
• Metabolic data collected from the following journals
•
•
•
•
•
•
•
Drug Metabolism and Disposition
Xenobiotica
Biochemical Pharmacology
Journal of Pharmacology and Experimental Therapeutics
Chemical Research in Toxicology
Journal of Medicinal Chemistry
Journal of Agriculture and Food Chemistry
Metabolism Database – Lasofoxifene example
C Prakash et al, Drug Metab. Dispos. 36 1218–1226 (2008)
Metabolism Database – Lasofoxifene example
Metabolism Database – Lasofoxifene example
Metabolism Database – Lasofoxifene example
C Prakash et al, Drug Metab. Dispos. 36 1218–1226 (2008)
Metabolism Database – Lasofoxifene example
E00
E00
Metabolism Database – Lasofoxifene example
E01
E00
E01
E00
Metabolism Database – Lasofoxifene example
O
O
O
O
O
O
O
O
H
H
N
O
N
O
O
H
H
E00
E01
Metabolism Database – Lasofoxifene example
• q = query compound (i.e. the parent compound, E00)
• o = observed experimentally (exact structure)
• p = presumed metabolite (authors describe as an exact structure but provide no evidence OR is a
hypothetical intermediate not observed e.g. alcohol to aldehyde to acid where aldehyde is presumed)
• x = expert call (ambiguous structure)
Metabolism Database – Lasofoxifene example
Metabolism Database – Data Diversity
Oxidation at Aromatic
Nitrogen
2%
N-Acetylation of Amines
2%
Carbonyl Reduction
2%
O-Dealkylation
2%
Conjugation with Amino
Acids
O-Methylation
2%
2%
Other Redox Reactions
3%
N-Glucuronidation of Amine
3%
N-Demethylation
3%
Carboaliphatic
Hydroxylation
22%
Miscellaneous Oxidative
Ring Opening
3%
Amide Hydrolysis
3%
O-Glucuronidation of
Alcohol
16%
O-Glucuronidation of
Carboxylic Acid or
Derivative
4%
O-Demethylation
4%
Dehydrogenation
5%
N-Dealkylation
5%
Carboaromatic
Hydroxylation
12%
O-Sulphonation
5%
Occurrence Ratio Method
How often does a reaction
actually occur?
Large
metabolism database
Occurrence Ratio
How often could a reaction
occur?
Occurrence Ratio Method: Biotransformation 243
How often does a reaction
actually occur?
612
Occurrence Ratio
0.314
How often could a reaction
occur?
1946
Occurrence Ratio Method – Making a Prediction
O
O
Rank using
Occurrence Ratios
N
Biotransformation
Number
27
243
253
20
78
66
69
69
67
245
235
100
66 243 20
67 235
78 69
100 245 27 253
Name
Score*
Glucuronidation of Aromatic Alcohols
Oxidative N-Dealkylation
Oxidative O-Dealkylation
O-Sulphonation of Aromatic Alcohols
Para Hydroxylation of Monosubstituted Benzene Compounds
Hydroxylation of Alicyclic Methylene Adjacent to an Aromatic Ring
Hydroxylation of Unfunctionalised Alicyclic Methylene
Hydroxylation of Unfunctionalised Alicyclic Methylene
Lactams from Aza-Alicyclic Compounds
Oxidative N-Dealkylation
5-Hydroxylation of 1,2,4-Trisubstituted Benzenes
Amine Oxides from Tertiary Alicyclic Amines
449
314
237
232
222
221
104
104
104
93
72
68
*Score = Occurrence Ratio x 1000
Occurrence Ratio Method – Scoring Filter
Top N Threshold (N=10)
•
O
O
Only display biotransformations
with the top N scores (N = 10)
N
Biotransformation
Number
27
243
253
20
78
66
69
69
67
245
235
100
Name
Score
Glucuronidation of Aromatic Alcohols
Oxidative N-Dealkylation
Oxidative O-Dealkylation
O-Sulphonation of Aromatic Alcohols
Para Hydroxylation of Monosubstituted Benzene Compounds
Hydroxylation of Alicyclic Methylene Adjacent to an Aromatic Ring
Hydroxylation of Unfunctionalised Alicyclic Methylene
Hydroxylation of Unfunctionalised Alicyclic Methylene
Lactams from Aza-Alicyclic Compounds
Oxidative N-Dealkylation
5-Hydroxylation of 1,2,4-Trisubstituted Benzenes
Amine Oxides from Tertiary Alicyclic Amines
449
314
237
232
222
221
104
104
104
93
72
68
Occurrence Ratio Method – Scoring Filter
Relative Threshold (50%)
•
O
O
N
Biotransformation
Number
27
243
253
20
78
66
69
69
67
245
235
100
Only display biotransformations
with scores at or above some
percentage of the maximum score
(e.g. 50% of 449 = 225)
Name
Score
Glucuronidation of Aromatic Alcohols
Oxidative N-Dealkylation
Oxidative O-Dealkylation
O-Sulphonation of Aromatic Alcohols
Para Hydroxylation of Monosubstituted Benzene Compounds
Hydroxylation of Alicyclic Methylene Adjacent to an Aromatic Ring
Hydroxylation of Unfunctionalised Alicyclic Methylene
Hydroxylation of Unfunctionalised Alicyclic Methylene
Lactams from Aza-Alicyclic Compounds
Oxidative N-Dealkylation
5-Hydroxylation of 1,2,4-Trisubstituted Benzenes
Amine Oxides from Tertiary Alicyclic Amines
449
314
237
232
222
221
104
104
104
93
72
68
CHEMICALLY INTELLIGENT
PREDICTIONS
Query-specific Occurrence Ratios
• Site of Metabolism (SOM) + Atom-centered fingerprint
O
1
N
N
O
N
N
BT242
N
N
O
N
N
O
2
N
2
Atom
feature
extractor
SOM 2: C; C-N; C-N-C; C-N-C=C.
Hashing
algorithm
1
0
0
1
0
Query-specific Occurrence Ratios
• Similar supporting examples (Tanimoto coefficient)
O
1
N
N
O
N
N
N
BT242
N
N
O
N
O
2
N
N
O
N
N
–
N
N
O
O
O
+
N
N
N
N
O
N
N
N
N
O
O
N
N
O
N
N
N
Query-specific Occurrence Ratios
• Similar supporting examples (Tanimoto coefficient)
O
1
N
N
O
N
N
N
BT242
N
N
O
N
O
2
O
N
O
S
N
N
N
N
O
N
N
N
O
S
N
N
O
N
N
N
O
N
N
O
N
N
O
EXAMPLES
Rotigotine – Meteor Nexus “Traditional Reasoning”
N
O
W Cawello et al, Drug Metab. Dispos. 37 2055–2060 (2009)
S
Rotigotine – Meteor Nexus “Traditional Reasoning”
N
O
First generation metabolites displayed.
S
Rotigotine – Experimentally Observed SOMs
N
S
O
O-Glucuronidation
O-Sulphonation
N-Dealkylation (thienylethyl)
N-Dealkylation (propyl)
Rotigotine – Experimentally Observed SOMs
N
O
W Cawello et al, Drug Metab. Dispos. 37 2055–2060 (2009)
S
Rotigotine – Meteor Nexus “SOM Reasoning”
N
O
S
Rotigotine – Meteor Nexus “SOM Reasoning”
N
O
S
Rotigotine – Meteor Nexus “SOM Reasoning”
N
O
S
PF-4991532 – Meteor Nexus “Traditional Reasoning”
F
O
N
F
F
N
N
N
O
O
Sharma et al, Drug Metab. Dispos. 43 190-198 (2015)
PF-4991532 – Meteor Nexus “Traditional Reasoning”
F
O
N
F
F
N
N
N
O
O
Sharma et al, Drug Metab. Dispos. 43 190-198 (2015)
PF-4991532 – Meteor Nexus “Traditional Reasoning”
F
O
N
F
F
N
N
N
O
O
Sharma et al, Drug Metab. Dispos. 43 190-198 (2015)
PF-4991532 – Experimentally Observed SOMs
Sharma et al, Drug Metab. Dispos. 43 190-198 (2015)
PF-4991532 – Meteor Nexus “SOM Reasoning”
F
O
N
F
F
N
N
N
O
O
Sharma et al, Drug Metab. Dispos. 43 190-198 (2015)
PF-4991532 – Meteor Nexus “SOM Reasoning”
F
O
N
F
F
N
N
N
O
O
Sharma et al, Drug Metab. Dispos. 43 190-198 (2015)
Terbinafine – Meteor Nexus “SOM Reasoning”
N
Relative Threshold = 50%
Biotransformation
Number
Biotransformation Name
Score
Intermediates
242
082
243
243
243
243
243
243
Oxidative N-Demethylation
Dihydrodiols via Arene Oxides
Oxidative N-Dealkylation
Oxidative N-Dealkylation
Oxidative N-Dealkylation
Oxidative N-Dealkylation
Oxidative N-Dealkylation
Oxidative N-Dealkylation
921
630
565
565
565
543
543
543
I22o
I2a
I7o
I10o, I11a
I14o, I15a
I6o
I8o, I9a
I12o, I13a
Terbinafine – Meteor Nexus “SOM Reasoning”
N
SL Iverson et al, Chem. Res. Toxicol. 14 175-181 (2001)
Cyclosporine G – Meteor Nexus “Traditional Reasoning”
O
O
O
N
N
N
N
N
O
O
O
O
N
O
O
N
N
N
N
O
N
O
JB Mangold et al, Drug Metab. Dispos. 22 873-879 (1994)
O
Cyclosporine G – Meteor Nexus “SOM Reasoning”
O
O
O
N
N
N
N
N
O
O
O
O
N
O
O
N
N
N
N
O
N
O
O
Scoring = Relative threshold, Score Threshold = 70%
Biotransformation
Number
071
075
242
Biotransformation Name
Score
Allylic Hydroxylation
Hydroxylation of Alkyl Methine
Oxidative N-Demethylation
473, 155
261, 249, 243, 222
261, 248, 245, 222
JB Mangold et al, Drug Metab. Dispos. 22 873-879 (1994)
Summary
• A hybrid machine-learnt approach to rank expert systemgenerated metabolites based on known metabolic reactions
reduces the number of “false positive” predictions
• Expert knowledge in biotransformation dictionary retained
• Time savings as manual biotransformation ranking rules no longer
needed
Summary
• A hybrid machine-learnt approach to rank expert systemgenerated metabolites based on known metabolic reactions
reduces the number of “false positive” predictions
• Further improvements can be anticipated as access to metabolic
reaction data is optimised
• ~17k reactions for ~2000 compounds manually curated from
published literature
Acknowledgements
• Chris Barber
• Tony Long
• Matt Wright
• Ed Rosser
• Catherine O'Leary-Steele
• Tim Furze
• Carol Marchant
• Emma Hill
• Jing Ma
• Jonathan Vessey
• Gila Scherer
• Lucy Gordon-Smith
• Joseph Dresner-Barnes
• Liam Corley
• Robert Davies
Thank You For Listening
Questions?