Will we ever be able to accurately predict solubility?

Accurate prediction of thermodynamic solubility by machine learning remains a challenge. Recent models often display good performances, but their reliability may be deceiving when used prospectively. This study investigates the origins of these discrepancies, following three directions: a historical perspective, an analysis of the aqueous solubility dataverse and data quality. We investigated over 20 years of published solubility datasets and models, highlighting overlooked datasets and the overlaps between popular sets. We benchmarked recently published models on a novel curated solubility dataset and report poor performances. We also propose a workflow to cure aqueous solubility data aiming at producing useful models for bench chemist. Our results demonstrate that some state-of-the-art models are not ready for public usage because they lack a well-defined applicability domain and overlook historical data sources. We report the impact of factors influencing the utility of the models: interlaboratory standard deviation, ionic state of the solute and data sources. The herein obtained models, and quality-assessed datasets are publicly available.

Synthesis of sulfaphenazole derivatives and their use as inhibitors and tools for comparing the active sites of human liver cytochromes P450 of the 2C subfamily.

Twenty-three new derivatives of sulfaphenazole (SPA) were synthesized to further explore the topology of the active sites of human liver cytochromes P450 of the 2C subfamily and to find new selective inhibitors of these cytochromes. These compounds are derived from SPA by replacement of the NH(2) and H (of the SO(2)NH function) substituents of SPA with various R(1) and R(2) groups, respectively. Their inhibitory effects were studied on recombinant CYP 2C8, 2C9, 2C18, and 2C19 expressed in yeast. High affinities for CYP 2C9 (IC(50) < 1 microM) were only observed for SPA derivatives having the SO(2)NH function and a relatively small R(1) substituent (R(1) = NH(2), CH(3)). Any increase in the size of R(1) led to a moderate decrease of the affinity, and the N-alkylation of the SO(2)NH function of SPA to a greater decrease of this affinity. The same structural changes led to opposite effects on molecular recognition by CYP 2C8 and 2C18, which generally exhibited similar behaviors. Thus, contrary to CYP 2C9, CYP 2C8 and 2C18 generally prefer neutral compounds with relatively large R(1) and R(2) substituents. CYP 2C19 showed an even lower affinity for anionic compounds than CYP 2C8 and 2C18. However, as CYP 2C8 and 2C18, CYP 2C19 showed a much better affinity for neutral compounds derived from N-alkylation of SPA and for anionic compounds bearing a larger R(1) substituent. One of the new compounds (R(1) = methyl, R(2) = propyl) inhibited all human CYP 2Cs with IC(50) values between 10 and 20 microM, while another one (R(1) = allyl, R(2) = methyl) inhibited all CYP 2Cs except CYP 2C9, and a third one (R(1) = R(2) = methyl) inhibited all CYP 2Cs except CYP 2C8. Only 2 compounds of the 25 tested derivatives were highly selective toward one human CYP 2C; these are SPA and compound 1 (R(1) = CH(3), R(2) = H), which acted as selective CYP 2C9 inhibitors. However, some SPA derivatives selectively inhibited CYP 2C8 and 2C18. Since CYP 2C18 is hardly detectable in human liver, these derivatives could be interesting molecules to selectively inhibit CYP 2C8 in human liver microsomes. Thus, compound 11 (R(1) = NH(2), R(2) = (CH(2))(2)CH(CH(3))(2)) appears to be particularly interesting for that purpose as its IC(50) value for CYP 2C8 is low (3 microM) and 20-fold smaller than those found for CYP 2C9 and 2C19.

Diclofenac and its derivatives as tools for studying human cytochromes P450 active sites: particular efficiency and regioselectivity of P450 2Cs.

A comparison of the oxidations of diclofenac with microsomes of yeasts expressing various human liver cytochromes P450 showed that P450 2C9 regioselectively led to 4'-hydroxy diclofenac (4'-OHD) whereas P450 3A4 only led to 5-hydroxy diclofenac (5-OHD). P450 2C19, 2C18, and 2C8 led to the simultaneous formation of 4'-OHD and 5-OHD (respective molar ratios of 1.3, 0.37, and 0.17), and P450 1A1, 1A2, 2D6, and 2E1 failed to give any detectable hydroxylated metabolite under identical conditions. P450 2C9 was found to be much more efficient for diclofenac hydroxylation than all the other P450s tested (k(cat)/K(M) of 1.6 min(-1) microM(-1) instead of 0.025 for the second more active P450), mainly because of markedly lower K(M) values (15 +/- 8 instead of values between 170 and 630 microM). Oxidation of diclofenac with chemical model systems of cytochrome P450 based on iron porphyrin catalysts exclusively led to the quinone imine derived from two-electron oxidation of 5-OHD, in an almost quantitative yield. Two derivatives of diclofenac lacking its COO(-) function were then synthesized; their oxidation by recombinant human P450 2Cs always led to a major product coming from their 5-hydroxylation. Substrate 2, which derives from reduction of the COO(-) function of diclofenac to the CH(2)OH function, was studied in more detail. All the P450s tested (1A1, 1A2, 2C8, 2C9, 2C18, 2C19, 2D6, and 3A4) almost exclusively led to its 5-hydroxylation. P450s of the 2C subfamily were found to be the most efficient catalysts for this reaction, with k(cat)/K(M) values between 0.2 and 1.6 min(-1) microM(-1). Oxidation of 2 with an iron porphyrin-based chemical model of cytochrome P450 also led to a product derived from the oxidation of 2 at position 5. These results show that oxidation of diclofenac and its derivative 2, either with chemical model systems of cytochrome P450 or with recombinant human P450s, generally occurs at position 5. This position, para to the NH group on the more electron-rich aromatic ring of diclofenac derivatives, is thus, as expected, the privileged site of reaction of electrophilic, oxidant species. The most spectacular exception to this chemoselective 5-oxidation of diclofenac derivatives was found for oxidation of diclofenac itself with P450 2C9 (and P450 2C19 and 2C18 to a lesser extent), which only led to 4'-OHD. A likely explanation for this result is a strict positioning of diclofenac in the P450 2C9 active site, via its COO(-) function, to completely orientate its hydroxylation toward position 4', which is not chemically preferred. P450 2C19, 2C18, and 2C8 would not lead to such a strict positioning as they give mixtures of 4'-OHD and 5-OHD. The above results show that diclofenac derivatives are interesting tools to compare the active site topologies of human P450 2Cs.

Comparison of the substrate specificities of human liver cytochrome P450s 2C9 and 2C18: application to the design of a specific substrate of CYP 2C18.

A series of 2-aroylthiophenes derived from tienilic acid by replacement of its OCH2COOH substituent with groups bearing various functions have been synthesized and studied as possible substrates of recombinant human liver cytochrome P450s 2C9 and 2C18 expressed in yeast. Whereas only compounds bearing a negative charge acted as substrates of CYP 2C9 and were hydroxylated at position 5 of their thiophene ring at a significant rate, many neutral 2-aroylthiophenes were 5-hydroxylated by CYP 2C18 with kcat values of >2 min-1. Among the various compounds that were studied, those bearing an alcohol function were the best CYP 2C18 substrates. One of them, compound 3, which bears a terminal O(CH2)3OH function, appeared to be a particularly good substrate of CYP 2C18. It was regioselectively hydroxylated by CYP 2C18 at position 5 of its thiophene ring with a KM value of 9 +/- 1 microM and a kcat value of 125 +/- 25 min-1, which are the highest described so far for a CYP 2C. A comparison of the oxidations of 3, by yeast-expressed CYP 1A1, 1A2, 2C8, 2C9, 2C18, 2C19, 2D6, 2E1, 3A4, and 3A5, showed that only CYP 2C8, 2C18, and 2C19 were able to catalyze the 5-hydroxylation of 3. However, the catalytic efficiency of CYP 2C18 for that reaction was considerably higher (kcat/KM value being 3-4 orders of magnitude larger than those found for CYP 2C8 and 2C19). Several human P450s exhibited small activities for the oxidative O-dealkylation of 3. The four recombinant CYP 2Cs were the best catalysts for that reaction (kcat between 1 and 5 min-1) when compared to all the P450s that were tested, even though it is a minor reaction in the case of CYP 2C18. All these results show that compound 3 is a new, selective, and highly efficient substrate for CYP 2C18 that should be useful for the study of this P450 in various organs and tissues. They also suggest some key differences between the active sites of CYP 2C9 and CYP 2C18 for substrate recognition.

Drug-induced immunotoxicity.

Immune-related drug responses are one of the most common sources of idiosyncratic toxicity. A number of organs may be the target of such reactions; however, this review concentrates mostly on the liver. Drug-induced hepatitis is generally divided into two categories: acute hepatitis in which the drug or a metabolite destroys a vital target in the cell; immunoallergic hepatitis in which the drug triggers an adverse immune response directed against the liver. Their clinical features are: a) low frequency; b) dose independence; c) typical immune system manifestations such as fever, eosinophilia; d) delay between the initiation of treatment and onset of the disease; e) a shortened delay upon rechallenge; and f) occasional presence of autoantibodies in the serum of patients. Such signs have been found in cases of hepatitis triggered by drugs such as halothane, tienilic acid, dihydralazine and anticonvulsants. They will be taken as examples to demonstrate the recent progress made in determining the mechanisms responsible for the disease. The following mechanisms have been postulated: 1) the drug is first metabolized into a reactive metabolite which binds to the enzyme that generated it; 2) this produces a neoantigen which, once presented to the immune system, might trigger an immune response characterized by 3) the production of antibodies recognizing both the native and/or the modified protein; 4) rechallenge leads to increased neoantigen production, a situation in which the presence of antibodies may induce cytolysis. Toxicity is related to the nature and amount of neoantigen and also to other factors such as the individual immune system. An effort should be made to better understand the precise mechanisms underlying this kind of disease and thereby identify the drugs at risk; and also the neoantigen processes necessary for their introduction into the immune system. An animal model would be useful in this regard.