Impact of Physicochemical Properties on Oral Drug Dose & Hepatotoxicity

Transcription

1 Impact of Physicochemical Properties on Oral Drug Dose & Hepatotoxicity Paul Leeson Paul Leeson Consulting Ltd UKQSAR, Syngenta, 26-Apr-217 1

2 Oral Dose: The Ultimate Composite Property Human dose size & frequency prediction is a central element of lead optimisation Dose driven by MEC/potency, F abs (linear); CL, V ss (exponential) High dose/exposure linked to preclinical & idiosyncratic toxicity including hepatoxicity (drug induced liver injury, DILI) Dose : solubility ratio >5 ml predicts formulation issues How do physicochemical properties affect: 1) oral dose & 2) DILI severity? Human dose prediction: Grime et al, Mol. Pharmaceutics, 213, 1, 1191; Dose/solubility: Butler & Dressman J. Pharm. Sci., 21, 99, 494: Bayliss et al, Drug Discov. Today 216, 21, 1719; Toxicity: Hughes et al, Bioorg. Med. Chem. Lett. 28, 18, 4872; DILI: Chen et al, Drug Discov. Today, 216, 21, 648; Hepatology 213, 58, 388 2

3 More % Oral drugs Oral Drug Doses: Compilation & Distribution Defined daily dose, DDD (n=1261), WHO: assumed average maintenance dose per day for a drug used for its main indication in adults. Many recent drugs lack a DDD Maximum daily doses, MDD (n=1841), highest dose recommended, or used for any indication. Oral drugs db updated to mid 216 & MDDs added -p[dose], not mg, used for SAR 2 Defined daily dose (DDD) Maximum daily dose (MDD) DDD n = 1261 MDD n = 1841 Mean -p[dose] % -p[dose] Median -p[dose] p[dose] bin 1% -p[dose] % dose (mg) 15 2 Median dose (mg) 1 2 1% dose (mg) Oral drug db: Leeson et al, Med. Chem. Comm. 211, 2, 91; MDD: and other prescription drug data worldwide; Chen et al, Hepatology 213, 58, 388; Stepan et al, Chem. Res. Toxicol. 211, 24, 1345; Matthews et al, Curr. Drug Discov. Technol., 24, 1, 61 (exclude non oral compounds); DDD: Weng et al, Oncotarget, 215, 6, ; pdose: Dearden et al, SAR and QSAR Environ. Res, 29, 2, 241 3

4 Oral Dose Reducing Over Time But significantly greater increase in physical properties Mean Mean Mol Wt Mean Mean clogp Mean Mean Mol Wt Mean Mean clogp -2.8 a d b c b c b 35 b b c c a c c a d c, d b b b, cb, c b, c 2.3 a, b c a c c c Publication decade Publication decade Points with different letters are statistically different (, p<.5) No significant change in dose since 197s Dose/property efficiency decreasing over time Oral drug physical properties vs time: Leeson & Davis J. Med. Chem 24, 47, 6338; Leeson & Springthorpe, Nat. Rev. Drug Disc. 27, 6, 881; Leeson et al, Med. Chem. Comm. 211, 2, 91 4

5 % Oral drugs 3 Dose & Ion Class Acid n=287 Base n=752 Neutral n=686 Zwitterion n= V d : Bases >Neutrals> Acids More bin Mean p[mdd] A -2.91; B -3.59; N -3.36; Z : A=Z; Z=N; B 5

6 clogp Oral Drug Lipophilicity vs, n=1841 Mean clogp Why? Mean clogp ± 2.5% MDD Highest 2.5% MDD Lowest 2.5% MDD Mean clogp* < -3.2 Lower 59% 21% > -2.47, Upper 2% * p<.1 Increasing clogp in oral drugs results in lower C free High dose high C free requires low clogp & high solubility Increased clogp & exposure linked to preclinical toxicity [C] free : Wenlock, Med. Chem. Comm. 216, 7, 76 & 1995; Toxicity: Hughes et al, Bioorg. Med. Chem. Lett. 28, 18,

7 Mean TPSA Mean TPSA/HA Mean Mol Wt Mean FSp3 Moving Means: Mol Wt, Fsp3, TPSA, TPSA/HA Mean Mol Wt ± 2.5% MDD Lowest 2.5% MDD Highest 2.5% MDD <-5.2 (lowest 5%) Mean Fsp3 ± 2.5% MDD Lowest 2.5% MDD Highest 2.5% MDD <-5.2 (lowest 5%) >-2.3 (highest 15%) Mean TPSA ± 2.5% MDD Lowest 2.5% MDD Highest 2.5% MDD Mean TPSA/HA ± 2.5% MDD Lowest 2.5% MDD Highest 2.5% MDD > -3. (highest 41%)

8 % Compounds 4/4 Rule & Dose Category L8% doses (n=1472) U2% doses (n=369) Company patents (n=792944) Oral drugs clogp <4 clogp <4 clogp >4 clogp >4 Mol Wt <4 Mol Wt >4 Mol Wt <4 Mol Wt >4 Oral drugs with Mol Wt>4 + clogp >4: 9.2 times more likely to have MDD in the lower 8% than in the upper 2% (odds ratio, p <.1) Candidates with clogp >4 & Mol Wt >4, the most exemplified category in patents, should be in L8% dose range ADME 4/4 rule: Gleeson, J. Med. Chem., 28, 51, 817 Patent data 2-11: Leeson & St-Gallay, Nat. Rev. Drug Disc. 211, 1, 749 8

9 9/369 Oral Drugs in U2% MDD have clogp >4 & Mol Wt >4; 3 Drugs, 3 Prodrugs & 3 Radiocontrast Agents Year is publication date Tiocarlide; tuberculosis; 8 mg (1953) Chloramphenicol palmitate; prodrug; antibiotic; 4 mg (1948) Iopanoic acid; radiocontrast medium; 6 mg (1955) Nelfinavir; HIV; 25 mg (1995) Talniflumate; prodrug; antiinflammatory; 22 mg (1978) Ipodate sodium; radiocontrast medium; 6 mg (196) Vemurafenib; late stage melanoma; 192 mg (211) Carbenicillin indanyl; prodrug; antibiotic; 4 mg (1969) Tyropanoate sodium; radiocontrast medium; 3 mg (1959) 9

10 # Oral drugs # Oral drugs Human Hepatoxicity Severity: Oral Drugs 116 Most-DILI & 48 No-DILI oral drugs from FDA label Rule of 2: highest DILI risk when clogp >3 & dose >1mg QSAR: difficult to understand physical basis; topology implicated DILIrank: 136 FDA approved drugs with annotated DILI severity 731 in oral db (others: injectables, topicals, biologicals, mixtures) Top 2 orals Oral drugs Median 1955 Median 1969 Publication decade 155 of top 2 by prescription are oral Reliance on historical drugs: is oldest class 6% of oral drug db have no DILI annotation Rule of 2: Chen et al, Hepatology 213, 58, 388; QSAR/topology: Xu et al, J. Chem. Inf. Model., 215, 55, 285 DILIRank: Chen et al, Drug Discovery Today, 216, 21, 648; Top 2: 1

11 DILI Class vs & clogp clogp Connected categories are not statistically different (p <.5, ) Means Means <, other categories are intermediate Consistent with rule of 2 11

12 DILI Sensitivity to clogp is Dose Dependent Consistent with overall lipophilicity vs dose trend Connected categories are not statistically different (p >.5, ) Means n=131 n=127 n=32 n=36 Mean clogp vs Lower 8% doses Δ clogp = 1. Upper 2% doses Δ clogp = No-DILI Most-DILI <-2.47 Lower 8% doses No-DILI Most-DILI >-2.47 Upper 2% doses 12

13 % Oral drugs DILI Class vs Fsp3 & Ion Class Connected categories are not statistically different (p <.5, ) Fsp3 Means Acid Base Neutral Zwitterion Fsp3: >, other categories intermediate No 3D differences (rod/sphere/disc) Reactive metabolites use sp2 centres Ion class influences DILI prevalence Neutrals: highest ; permeability? Acids: acylglucuronide formation? 32/44 acids are carboxyls Reactive metabolite chemistry: Stepan et al, Chem. Res. Toxicol. 211, 24,

14 # Oral drugs Acids: Lipophilicity & Fsp3 more important than dose n=59: 15; clogp Recursive Partition Fsp3 p = Mean values p < LogDpH LogD6.5 <-1.75 LogD6.5>-1.75 % Most p < Mean values -4.5 p < % No 8. Odds ratio (CI) 84. ( ) p <.1 14

15 # Oral drugs Bases: Dose & lipophilicity more important than Fsp p = Fsp3 p < n=129: 88; Mean values Mean values -.53 clogp p = LogDpH6.5 p = nd 1 st Recursive Partition 7 58 < LogD6.5 < LogD6.5 >-.26 > % Most 73.2 % No 86.4 Odds ratio (CI) 17.3 ( ) p <.1 15

16 # Oral drugs Neutrals: Dose & Fsp3 more important than lipophilicity p = Fsp3 p = n=12: 48; Mean values Mean values clogp p = LogDpH6.5 p = rd 2 nd 1 st Recursive Partition 3 19 < LogD6.5 < LogD6.5 >2.17 Fsp3 > > % Most 8.6 % No 85.4 Fsp3 <.29 Odds ratio (CI) 24.3 ( ) p <.1 16

17 All Compounds: Recursive Partition n=326: 163; 163 Data splits 3 rd clogp 2 nd Fsp3 1 st prediction prediction clogp <2.374 clogp >2.374 Fsp3 >.28 Fsp3 <.28 < > % Most 85.3 % No 79.1 Odds ratio (CI) 22. ( ) p <.1 LogDpH6.5 < or >2.2 (instead of clogp <>2.374) % Most 74.8 % No 89. Odds ratio (CI) 24. ( ) p <.1 17

18 Model Test Set: 21 Oral Drugs Discontinued in Development Due to Human Hepatotoxicity Compound Ref Max Daily Dose mg clogp Fsp3 DILI prediction Most vs No Model Fails rule of 2 (Dose >1 mg & clogp >3) CP a Most No CP a Most No CP a Most Fail CP a Most Fail Zamifenacin a Most No CP a Most Fail Darbufelone a No No CP a Most Fail TAK-875 a Most No LY a Most No MK-893 a Most Fail ADX-159 a Most Fail Telcagepant b Most Fail Sitaxentan c Most Fail Fialuridine d No No Fiduxosin d Most Fail Falnidamol d Most Fail Pafuramidine d Most Fail Tasosartan d Most No Pralnacasan d No No Aplaviroc d Most Fail % Predicted 86% (18/21) 57% (12/21) a. Shah et al Tox. Sci., 215, 147, 5 b. Ho et al, Neurology, 214, 83, 958 c. d. Chen, et al, Drug Discov. Today 216, 21,

19 Is Most/No DILI Model Useful Predictively? Model distinguishes oral human drug known DILI extremes (DILIRank) Uses well known properties: dose, Fsp3, lipophilicity, ion class Caveats: large number of uncategorised drugs & newer drugs in changed property space Mol Wt, clogp vs model DILI Class n Oral drugs published post-199 (n=247) RP Model Prediction n n Less DILI Ambiguous DILI No category Majority of recent drugs have no human DILI assignment Hepatotoxicity yet to appear? Use in conjunction with hepatotoxicity screens BSEP, reactive metabolites, covalent binding, cellular toxicities OK to ignore? (Top 2 = 59% Less DILI ) 19

20 Acknowledgements Scott Boyer (AZ) Dose-Properties Minjun Chen (FDA) DILIRank DB Mike Mortimore (Vertex) 3D Analysis 2

21 Upper 2% < Lower 8% Upper 2% > Lower 8% Upper 2% = Lower 8% Summary of Mean Properties Lower 8% vs Upper 2% MDD Mean Property Values Property < n = 1472 (lower 8%) > n = 369 (upper 2%) Difference p clogp <.1 LogD <.1 PFI <.1 Mol Wt <.1 Ar ring <.1 Ar ring/ha <.1 Positive charge <.1 O+N/HA <.1 OH+NH/HA <.1 TPSA/HA <.1 Negative charge <.1 Chiral atoms/ha Fsp sp2-sp

22 Mean Physical Properties by DILI Class N M < Less DILI L Ambiguous DILI A > No category Nc Categories in descending property order. Groups separated by ; are statistically different (Tukey, p<.5) n M; L=Nc=A; L=A=N clogp M=Nc; Nc=L; L=A=N LogD M=Nc; L=A; N LogD M=Nc; L=A=N LogP M=Nc; L=A=N PFI M=Nc; L=N=A Ar ring M=L; Nc=L=A; L=A=N sp M=Nc=L; Nc=L=A; A=N sp2-sp M=L; L=Nc=A; A=N sp N=A=Nc=L; A=N=L=M Chiral atoms N=L=A; L=A=Nc=M Csp N=Nc=A=L; A=L=M Fsp N=A; A=Nc=L; M HA No differences CMR No differences Mol Wt No differences RotB No differences HBA No differences HBD differences; A=L=M=Nc O+N No differences OH+NH differences; A=L=M=Nc TPSA No differences = 22