Table Ill. Predicted Results for log IV Using Neural Networks and
`Regre>sion Analysis
`
`J. Am. Chem. Soc. 1991, I 13, 9483-9493
`
`9483
`
`est( RA)&
`-1.61
`-2.03
`-5.28
`-0.44
`
`-5.66
`
`-4.49 "
`-3.58 "
`
`r<f
`II
`18
`18
`25
`
`26
`
`e>t(NN)a
`e~pt
`-[ .40
`I. 4-heptanol
`-1.40
`-2.72
`-2.35
`2. menthone
`I, 1 ·diphcnylethy!cne
`-5.23
`-4.52
`3.
`-0.53
`4. p·cresol
`-0.81
`testosterone
`-4.66
`-4.08
`5.
`6. 2,4,4'·PCB
`-6.24
`-5.96
`-3.47
`-3.59
`1. de~amethasone
`-1.83
`-2.85
`8. 4 ·ch I or on it robc:nzene
`II
`-2.66
`9. 2,5-PCB
`-5.45
`-5.06
`21
`-5.15
`-5.35
`JO. 2,6·PCB
`21
`-5.52
`-5.21
`-5.88
`-6.24
`-6.06
`11. 2,4,6·PCB
`21
`-4.78
`-4.92
`J 2. nuorene
`-4.43
`JO
`13. pyrene
`-6.04
`-6.17
`JO
`-6.39
`-3.27
`-3.10
`-3.04
`14. indan
`JO
`15. 3-methylpyridine
`--0.\ 1
`0.04
`II
`--0.01
`16. isoquinoline
`-I.I I
`-1.24
`II
`-J.45
`0,74
`11. tetrahydrofuran
`0.59
`0.48
`II
`-3.55
`-2.95
`-3.27
`18. cortisone
`25
`19. 2-nnphthol
`-2.08
`-2.25
`-1.61
`25
`~Estimation of tog IV using neural networks which gives a standard
`deviation 0.43. The standard deviation is 0.37 if we ka\'e out the 4.
`ch!oronitrobcnzene.
`! Estimation of log IV using regression analysis
`which gives a standard deviation 0.36.
`
`the neural network's predictive ability. Care should be taken in
`interpreting the results, however, since strictly the neural network
`should only be applied to predicting those compounds containing
`the particular substituents found in the training set. The results
`obtained arc shown in Table Ill together with the values predicted
`by the regression analysis technique. Again the performance of
`the neural network is very satisfactory and oompares favorably
`with that given by the regression analysis method. The neural
`network gives a predicted aqueous solubility superior to that
`obtained by regression analysis in 9 of the 19 cases. The poor
`value predicted for 4-ch!oronitrobenzcne is probably due to the
`omission from the training set of any ch!oronitro compound which
`would reduce the credence attached to the predicted value.
`In conclusion, a neural network model has been applied to the
`prediction of the aqueous solubility of organic compounds and
`the usefulness of the model clearly demonstrated. The predictive
`capabiHty of neural networks has been demonstrated on a number
`of unknown organic compounds. It has been shown in this study
`that neural networks give a superior performance to that given
`by a regression analysis t~hnique. \Vhi!e this work was in progress
`a paper was published34 describing an application of the neural
`network approach to estimating quantitative structure-activity
`relationships. This work confirms the conclusions derived in this
`study that neural networks can determine such relationships with
`a performance exceeding that of linear multiregression analysis.
`Clearly, the neural network approach would seem to ha\'C great
`potential for determining quantitative structure-activity rela·
`tionships and as such be a valuable tool for the metlicinal chemist.
`
`Supplementary J\'laterlal Al'ailable: Listing of complete ex·
`perirnental and estimated log W values (6 pages). Ordering
`information is given on any current masthead page.
`
`(34) Ao)·oma, T.; Suzuki, Y.; lchikaw3, H.J. ~ftd. Chem. 1990, JJ, 2583.
`
`found to be superior to that obtained with the regression analysis
`approach, 0,30. The results clearly demonstrate that the neural
`network has captured the association between the selected prop·
`ertics of an organic compound and its aqueous solubility.
`The trained neural network was tested on its ability to predict
`the aqueous solubility of an unknown set of organic compounds,
`that is, the compounds were not members of the original training
`set and indeed in some cases were quite unrelated to the original
`members. The test set should therefore provide a severe test of
`
`Computational Studies on FK506: Conformational Search and
`Molecular Dynamics Simulation in Water
`
`Julianto Pranata and \Villiarn L. Jorgensen*
`Contribution fron1 the Depart111ent of Che1nistry, Yale University, i\'ew Haven, Connecticut 0651 I.
`Received }lfay 6, 1991
`
`Abstract: Computational investigations have been undertaken to elucidate the conformational characteristics and the hydration
`of the immunosuppresant FK506. The calculations made use of the AMBER/OPLS molecular mechanics force field, augmented
`with some newly developed parameters particularly for the a·ketoamide torsion, A conformational search on FK506 using
`an internal coordinate ]'.fonte Carlo method found 21 distinct energy minima within 12 kcal/mo! of the lowest energy structure.
`The minima include structures with both cis and trans conformations of the amide bond. A 200·ps molecular dynamics simulation
`in water then provided information on the dynamical behavior of the cis isomer of FK506 as well as its hydration. Two
`oonform~tions of the macroc)'clic ring are sampled during the simulation. and some exocyc!ic groups undergo rapid conformational
`ch,1ngcs. Considerable flexibility is also observed near the amide functionality, which is in the binding region of FK506. The
`hydration of FK506 shows interesting variations owing to differences in the stcric environments of potential hydrogen-bonding
`In the critical binding region, there are on average 5 hydrogen bonds between water molecules and FK506.
`sit.:s.
`
`FK506, rapamycin, and cyclosporin A (CsA) are immuno(cid:173)
`supprcslve agents that act by blocking the signal transduction
`pathways that lead to T lymphocyte activation. 1 FK506 and
`rapamycin are structurally similar and appear to bind to the same
`receptor, FKBP,2 while the structurally unrelated CsA, a cyclic
`undecapeptide, binds to a different receptor, cyclophilin.1 Both
`
`(!)Schreiber, S. l. Science 1991. 151. 283.
`(2) Bierer, B. E.; Mattila, P. $.;Standaert, R. F.; Hcrzcnberg, L.A.;
`Burakoff, S. J.: Crabtree. G.; Schreiber, S. l. Pra<:. Natl. Acad. Sd. U.S.A.
`1990. 87, 9231. Dumont, F. J.: Melino, M. R.; Staruch, M. J.; Koprak, S.
`L.; Fischer, P.A.: Sigal, N. H.J. fmmunol. 199<1, /'14, !418. Fretz, H.;
`Albers, M. \V.; Galat, A.: Standaert, R. F.; Lane, \V. S.; BurJkoff, S. J.;
`Bierer, B. E.; Schreiber, S. I.. J. Am. Clum. Soc. 1991, 113, 1409.
`
`receptors have been shown to be peptidyl·prolyl cis-trans isom·
`erases (rotamascs). 4.5 FK506 and rapamycin inhibit the rotamase
`activity of FKBP, but not of cyclophilin: likewise CsA inhibits
`the rotamase activity of cyclophilin, but not of FKBP. 4
`
`(3) Handschumacher, R. E.: Harding, M. \V.; Rice, J.; Drugge, R. J.;
`Sp:icher, 0. \V, S,ience 1984, 226, 544. Handschumachcr, R. E.: Harding,
`M. \V, Tramplanta1io11 1988, 46, 29S.
`(4) Harding, M. \V.: Gala!, A.; Uehling, D. E.; Schreiber, S. l. Na1ure
`1989, 141, 758. Siekierka, J. J.; Hung, S. H. V.; Poe, M.; Lin, C. S.; Sigal,
`N. H. Nature 1989, 341, 155.
`(5) Fi~chtr, G.; \Vittmann·Licbold, B.: Lang, K.; Kiefhaber, T.; Schmid,
`F. X. Naturt 1989, 226, 544. Takahashi, N.: Hayano, T.: Suzuki, M. Nature
`1989, 226, 473.
`
`0002-7863/91/1513-9483$02.50/o
`
`© 1991 American Chentical Society
`
`Roxane Labs., Inc.
`Exhibit 1016
`Page 001
`
`

`
`9484 J. A1n. Chen1. Soc., Vol. I I 3, No. 25, 1991
`
`Pranata and Jorgensen
`
`20.0 - - - - - - - - - - - - - - - - - - -
`
`'ii s 15.0
`1
`i;t 10.0
`]
`1
`f
`"
`
`5.0
`
`. ·~.
`
`0.0
`
`50.0
`
`'20.0
`
`180.0
`
`OCCO O'hedrol Angle (degrees)
`Figure I. Torsional profiles of N,iV·dimethyl·ci:·ketopropanamide from
`molecular orbital calculations.
`
`limited the conformational sampling.
`\Ve are interested in investigating the structural and energetic
`aspects of the binding of immunosuppresants and substrates to
`FKBP. The initial efforts, described here, addressed the devel·
`opment of needed force-field parameters for FK506, conforma·
`tional search for low-energy structures in the absence of solvent,
`and characterization of the hydration and internal motions of
`FK506 through a 200-ps molecular dynamics simulation in water.
`Parameter DeYclopment
`A principal difficulty in performing computations on molecules
`like FK506 is the lack of appropriate molecular mechanics pa·
`rametcrs. For proteins and nucleic acids, a variety of standard
`parameter sets arc available, e.g., AMBER 16 or CHARMm. 11
`However, FK506 contains functionalities not found in peptides
`or nucleotides. In particular, an accurate description of the a·
`ketoamide torsion is required, in view of its importance in the
`binding process.11
`\Vhere available, parameters and potential functions from the
`A~iBER force field were used for bonded interactions (bond
`stretches, bends, and torsions, including improper torsions). 16 The
`OPLS parameters and functions were used for nonbonded in(cid:173)
`teractions.18 Some parameters were added to the AMBER set
`on the basis of existing paran1eters for similar functional groups.
`Appropriate parameters were not found for some torsions involving
`eroters, ketoncs, and o!efins. Parameters appropriate for an ester
`group were recently developed by Charif son et al.19 and were used
`in the present work. Other torsional parameters were obtained
`by fitting to ab initio torsional profiles calculated by \Vibcrg and
`co-workers.N A united-atom model has been here for CH,. units,
`otherwise all atoms are explicitly represented.
`As mentioned above, the a·ketoamide torsion is particularly
`important, and the parameters for this functionality were developed
`with the aid of quantum mechanical calculations. N,1Y-Di(cid:173)
`mcthyl·a·ketopropanamide was chosen as a n1odel system, and
`the torsional profile for the bond between the two carbonyl carbons
`was computed using scmiempirical (A/'.11)11.22 and ab initio
`
`(16) Weiner, S. J.; Kollman, P.A.; Case. D. A.: Singh, U. C.; Ghio, C.;
`A!agona, G.: Profeta. S .. Jr.: Weiner, P. J. Am. Chem. Soc. 19g4. /06, 165.
`\Veiner, S. J.; Kollman, P.A.; Nguyen, D. T.; Case, D. A. J. Comp. Chem.
`1986, 7. 230.
`(!7) Brooki, 8. R.; Bruccoleri, R. E.: Olafson, B. D.; States, D. J.; Sw·
`aminathan, S.; Karplus, M. J. Comp. Chttn. 19g3, 4, \g7, Nilsson, L.;
`Karplus, M. J. Comp. Chem. 1986, 7. 591.
`(Jg} Jorgensen, \V. L: Tirado-Ri~·e>, J. J. Am. Chem. Soc. 1988, 110,
`1657. Jorgensen, \V. L.; Rriggs, J. M.:Contcras, M. l. J. Phys. Chem. 1990,
`94, J6g3,
`(19} Charifaon, P. S.; Hiskey, R. G.; Pcdcr~en, l. G. J. Comp. Chem.
`1990, 10, 11g1.
`(20) \Vibcrg, K. B.; ~fartin, E. J. Am. Chem. Soc. 1985, 107, 5035.
`\Viberg. K. B. J. Am. Chem. Soc. 1986, 108, 5817. \Vibug, K. B.: Laidig,
`K. E. J. Am. Chem. Soc, 19g7, 109, 5935.
`(21) Dev.ar, M. J. S.; Zoebisch, E.G.; Heal)" E. F.: Stewart, J. J.P. J.
`Am. Chem. Sac. 1985, 107, 3902.
`(22) A~{l calculations were p<:rformed using the MOPAC program.
`Stev.art, J. J. P. MOPAC S.O; QCPE Program No. 455; Indiana University.
`Bloomington, JN.
`
`'~'\Jl)
`~t{: ,~o ~. 'f-y
`.. -;,:J'. (--,j)o
`_.(o.J__i'."(1,
`
`C'i~llA
`Crystal structures of all three immunosuppresants have been
`reported.6-8
`In addition, N~iR data have been used to deduce
`the structure of CsA in a variety of soh•ents,6.9 The solution-phase
`structure for FK506 in CDCh has also been examined. 10
`Schreiber and co-workers have investigated the inhibition of
`FKBP rotamase activity by FK506. 11 The immunosupprcsant
`with llC labels at C8 and C9 was synthesized 11 and used to probe
`the binding process. llC NMR of the enzyme-inhibitor complex
`shows no evidence for the existence of a tetrahedral adduct at
`either C8 or C9. Thus, the mechanism for rotamase activity does
`not involve the formation of a tetrahedral intermediate; rather,
`a mechanism was proposed which features a twisted amide bond
`in the transition state. The a·ketoamide functionality in FK506
`(and rapamycin) has an orthogonal orientation in the crystal
`structurc1•1 and serves as a surrogate for the twisted amide bond.
`Thus, by acting as a transition-state analogue, FK506 potently
`inhibits rotamase activity.
`Interestingly, the NMR of FK506 in solution shows the ex:·
`istence of two isomers, attributed to the cis and trans conformations
`for the amide bond in a 2:1 ratio. 10•11
`It has recently been de·
`tcrmined by X-ray crystallography that the trans isomer is bound
`to FKBP, 13 though only the cis isomer is observed in the crystal
`structure of isolated FK506. 1 For CsA, both in the crystal and
`in solution the molecule has a cis peptide bond b-etween residues
`Me Leu 9 and Me Leu I Q.6.9 However, there is evidence that CsA
`bound to cyclophilin also adopts a trans conformation for this
`bond.14
`On the computational side, Lautz et al. have reported molecular
`dynamics simulations of CsA in water, CCl4, and the crystalline
`environment. \j The focus was on comparisons with experimental
`structural data and medium effects on the conformation of CsA.
`However, the 40-50-ps durations for the simulations severely
`
`{6) Loosli, H. R.: Ke'islcr. H.; Oschkinat, H.; Weber, H.P.; Pelcher, T.
`J.; Widmer, A. Htlv. Chim. AC/a 1985, 68, 682.
`(7) Tanaka, H.: Kuroda, A.; Maruiawa, H.; Hatanaka, H.: Kino, T.; Goto,
`T.; Ha1himoto, M.; Taga, T. J. Am. Chem. Soc. 1987, 109, 5031. Taga, T.;
`Tanaka. H.; Goto, T.; Tada, S. Acta Crys1af/ogr, 1987, C</3, 151.
`(g) Swindell~. D. C. N.: \Vhite, P. S.: Findlay, J. A. Can. J. Chem. 1978.
`56. 2491.
`(9) Kessler, tl.; Oschkinat, H.: LOOili, H. R. Htlv. Chim. Ac1a 1985, 68,
`661. Kessler, H.; Ki.ick, M.: \Vein, T.; Gehrke, M. lftlv. Chim. Acta 1990,
`73, 1s1g,
`(10) Karuso, P.; Kc~sler. H.; ~fierke, D. F. J. Am. Chem. Sor. 1990, I I],
`9434.
`(11) Ros-!n, ~f. K.; Standaut, R. F.: Galat, A,; Nakatsuka, M.; Srhreibu,
`S. L. Sdtnct 1990, 1,8, 863.
`(12) Nakatsuka, M,; Ragan, J. A.; Sammakia, T.; Smith, D. B.; Uehling.
`D. E.; Scl1tciber, S. l. J. Am. Chtm. Sor. 1990, Ill, 5583.
`(13) Van Duyne, O. D.; Standaert, R. F.; Karp!us. P.A.; Schrcibu, S. L.;
`Clardy, J. Scitnct 1991, 151, 839.
`(14) fe>ik, S. \V.: Gampe, R. T., Jr.; Holzman, T. F.: Egan. D. A.: Edalji.
`R.; Lui)', J. R.; Simmer. R.: Helfrich, R.; Kishore, V,; Rich, D. H. Scfrnce
`1990, 150, 1406.
`( l S) Lautz, J.; Ke5sler, H.; van Gunstercn, \V, f.; Weber, H. P.; \Vengcr,
`R. M. Blopo!ymus 1990, 19, 1669. Lautz, J.; Keuler, A.; Kaptein, R.; van
`Ounuercn, \V. F. J. Compultr·Alded Mo!. Dtslgn 1987, /, 219.
`
`Roxane Labs., Inc.
`Exhibit 1016
`Page 002
`
`

`
`Computational S111dies 011 f'K506
`
`J. A111. Chem. Soc., Vol. 113, No. 25, 1991 9485
`
`Table I. New AMBER Torsion Parameters
`torsion
`
`CH-C(=0)-0-C•
`o=c-o-c~
`0-C(=O)-CH-X
`
`C-C(--0)-C-X
`C-C(~OJ-CH-X
`C-C(=0)-CH 1-X
`
`x-c~c-x
`C=CH-CH1-X
`c~cH-CH-X
`C=C-CH 1-X
`c~c-CH-X
`C-C(~C}-CH,-X
`C-C(~C}-CH-X
`
`o~c-c~o
`O=C-C-N
`C-C-C=O
`C-C-C-N
`•Reference 19.
`
`Vif2
`
`2.54S
`
`0.470
`
`0.203
`0.30l
`0.610
`
`0.700
`O.JlO
`0.3l0
`0.17l
`0.350
`0.175
`
`0.12
`0.12
`0.12
`0.12
`
`>
`
`0.0
`
`0.0
`
`0.0
`0.0
`0.0
`
`Ester
`
`Ketone
`
`Olefin
`
`Vlf2
`
`4.!50
`o.soo
`
`0.230
`0.34S
`0.690
`
`7.500
`
`180.0
`180.0
`180.0
`180.0
`0.0
`0.0
`
`«·Ketoamide
`0.42
`180.0
`0.42
`0.0
`0.42
`0.0
`0.42
`180.0
`
`>
`
`180.0
`180.0
`
`180.0
`180.0
`180.0
`
`l80.0
`
`180.0
`180.0
`180.0
`180.0
`
`V3/2
`
`l.100
`O.llO
`O.llO
`0.275
`0.550
`0.275
`
`180.0
`180.0
`180.0
`180.0
`0.0
`0.0
`
`(HF /l-21 G and HF /6-J I G(d)/ /l-21 GJ'"" molecular orbital
`calculations. The results arc shown in Figure !. Both the AMI
`and 6-JJG(d) calculations predict a nonplanar minimum, although
`the barrier at the anti conformation is quite low. For AMI. the
`minimum occurs at 124° and the height for the anti barrier is
`0. 79 kcal/mo!, while the corresponding values from the 6-31 G(d)
`calculations nre !35° and 0.65 kcal/mo!. However, the 3-2\G
`results appear to overestimate the stability of the anti confor(cid:173)
`mation, making il the minimum. In the crystal structures, the
`a·ketoamide functionality in both FK506 and rapamycin has an
`orthogonal orientation.7•8 The same trend is observed in other
`molecules when this functionality is doubly substituted at the
`nitrogen; e.g., tetramethyloxamide is twisted by 7 l 0 ,is
`ANI BER-type parameters were then obtained by fitting to the
`6-3lG(d) torsional profile. The fitting took into account the
`substantial nonbonded interactions in this molecule. In fact, it
`turned out that the nonbonded (i.e., steric) interactions are solely
`responsible for the nonplanarity of the system: there is nothing
`unusual ab-Out the torsional parameters that are reported in Table
`I.
`
`As a test of the new parameters, they were incorporated into
`Ar-.1BER and used to compute the torsional profile of a non(cid:173)
`macrocyc!ic fragment of the FK506 structure. This is compared
`to the profile computed using A!\-11 in Figure 2. Although the
`Ar-.iBER profile has a shallower minimum, the difference is not
`great, and the two profiles have the same overall shape. The
`minimum in the results with AMBER occurs at -108°, compared
`to ~96° with AM I, while in the crystal structure of FK506 this
`torsional angle is -89° .7
`All of the torsional parameters added to the AMBER set for
`the purpose of this work are presented in Table I. In addition,
`a complete listing of the parameters for FK506 is given in the
`Supplementary Material. The same parameters were used for
`the conformational search and molecular dynamics.
`Conformation Search
`Procedure. A fairly extensive search for the conformational
`minima of FK506 in the ideal gas phase was performed using an
`
`{23) Hchrc, W. J.; Radom, L.: Schleycr, P. v. R.; Pople, J. A. Ab lnitio
`J.fofecular Orbital Tlitor)'; \Viley: New York, 1986.
`{24) Ab inilio calculations were performed using the GAUSSIAN 90
`programs. Frisch, M. J.: Hcad·Gordon, M.: Trucks. 0. W.; Foresman, J.B.;
`Schlegel, H. B.; Ra8havachati. K.: Robb, M.A.: Binkley, J. S.: Gonzalez. C.:
`Dcfrccs, D. J.; Fox, D. J.: \Vhitcside, R. A.; Steger, R.; Melius, R.: Baker,
`J.: Martin, R. L.: Kahn, l. R.; Stewa1t, J. J.P.: Topiol, S.; Poplc, J. A.
`GAUSSIAN 90 Revision F: Gau55ian Inc.: Pittsburgh, PA, L990.
`(25) Adiv.idjaja, G.: Yoss, J. Cliatt. Bu. 1977, J 10, 1159.
`
`10.0 ~----~-----------,
`
`, ,
`
`/
`
`/
`
`0,,)-._ .:_,l
`
`1~~u
`
`1
`~ 5.0.
`!!.
`
`-5.o L ___ -2-====:::-_~--~__J
`-60.0
`-180.0
`-120.0
`0.0
`
`OCCO Dihedr<JI Arigle (degrees)
`Figure 2. Torsional profiles of a fragment of FK506.
`
`internal coordinate Monte Carlo method. 26 The focus was on
`the conformation of the 2l·membered macrocycle. No search
`was conducted which involved variations of the exocyclic torsions.
`Also excluded were torsions within six-membered rings, namely
`those for the C2-N7, C!o--05, and 05-C\4 OOnds, as well as the
`double bond (C!9-C20). The remaining 17 dihedral angles were
`randomly varied; however, 01---CI was defined as the ring-closure
`bond, so its lorsion and the torsions around the two adjacent bonds
`(Cl-C2 and C26-01) were not explicitly varied. Of course, no
`conslraints were applied in the subsequent energy minimizations.
`The starting structure for the search was obtained from the
`X-ray crystal structure.1 Initially, all hydrogens bound to carbons
`were removed, resulting in representation of FK506 as 60 explicit
`atoms. This structure was energy minimized, and the resulting
`structure was used to start the conformational search. The R~fS
`deviation between the actual X-ray structure and the energy·
`minimized form is only 0.48 A. The only significant change is
`the formation of a hydrogen bond between 06-H l and 04 that
`is not in the crystal structure (vidc infra). A total of 8499
`structures were generated using a random walk procedure: these
`were initially minimized to a root mean square gradient of l
`kJ/(mol·A) (=0.239 kcal/(mol·A)). Nonduplicate structures
`whose energies were within 50 kJ/mol (= 12 kcal/mo\) above the
`lowest energy minimum were saved. This resulted in 28 structures
`which were further minimized to a root mean square gradient of
`
`(26) Chang, G.: Guida, \V. C.: Still, \V. C. J. Am. Chem. Sor:. 1989, 111,
`4379.
`
`Roxane Labs., Inc.
`Exhibit 1016
`Page 003
`
`

`
`9486 J. A1n. Cheni. Soc., Vol. I J 3, No. 25, J 991
`
`Pranata and Jorgensen
`
`Table ll. Values of Macrocyc\ic Dih«lral Angles (deg} in the Conformations Found in the Monte Carlo Search and in the X-ray Structure
`conformation
`5
`dihedral angle
`8
`7
`4
`2
`3
`6
`-179
`-I 78
`87
`-168
`91
`179
`86
`43
`0JC1C2N7~
`-100
`-100
`-93
`-95
`-99
`-10!
`-95
`CIC2N7C8'
`-100
`-2
`-3
`-5
`0
`165
`2
`I
`0
`C2N7C8C9
`-87
`-99
`-63
`-96
`-68
`-1l5
`-104
`167
`- !OS
`N7C8C9C!O
`68
`81
`68
`91
`65
`90
`75
`C8C9C\005
`55
`72
`177
`175
`179
`178
`173
`179
`178
`177
`169
`C9CJ005CJ4'
`-!77
`-179
`-177
`-l 77
`-175
`-177
`-176
`-176
`-176
`C!005Cl4CIS'
`76
`75
`78
`62
`86
`75
`05C\4Cl5Ci6
`49
`85
`92
`68
`65
`65
`57
`Cl4Ci5Cl6Cl7
`-78
`89
`68
`6Q
`52
`-159
`-159
`-170
`-169
`-172
`-168
`-\Sl
`-179
`-170
`Cl5C\6Cl7Cl8
`70
`-63
`167
`-56
`60
`155
`66
`60
`C!6CI7CISCl9
`74
`-121
`-126
`-123
`-107
`71
`Ci7C\8C19C20
`-119
`-67
`16
`130
`-176
`-179
`-177
`175
`176
`-176
`179
`171
`CJ8C!9C20C21'
`l7l
`-142
`-123
`-118
`-124
`-124
`-115
`-115
`-I 10
`C19C20C2!C22
`-129
`c2oc21c22c21
`-48
`ll8
`-39
`83
`106
`JOI
`7J
`Ill
`IOI
`-99
`-157
`148
`-148
`-120
`c21c22c23c24
`174
`122
`178
`-152
`-65
`J 75
`-64
`169
`169
`146
`-58
`C22C2JC24C25
`172
`172
`-84
`-71
`-65
`-69
`-163
`-69
`-61
`-145
`-67
`C23C24C25C26
`-2)
`-56
`-24
`-37
`-55
`45
`-42
`-40
`C24C25C2601
`-23
`-154
`-149
`-147
`-147
`-153
`-146
`-147
`-!55
`-154
`C25C260lCJb
`-160
`-162
`-169
`-176
`-169
`-111
`-171
`-165
`-175
`C260lCIC2b
`25.9
`25.7
`24.5
`24.4
`23.9
`23.3
`20.6
`17.6
`energyJ
`26.4
`~After minimization. bQJCI is defined as the ring closure bond, thus these dihedral angles are not explicitly included in the search.
`
`9
`-165
`-100
`
`-·
`
`'These
`
`B I
`
`i l 1
`
`"
`
`I
`
`r
`
`i
`I
`9
`
`~
`
`~ v
`v
`~ "'
`v
`'O
`~
`v
`
`-'" 0 >
`
`180.0
`
`90.0
`
`0.0
`
`-90.0
`
`-180.0
`
`I
`
`i
`
`j
`
`!
`
`I
`
`6
`
`!
`
`11
`
`16
`
`21
`
`Dihedral Angle
`Figure J. Distribution of macrocydic dihedral angles in the 2l structures found in the ~fonte Carlo search. Open squares are for trans amide isomers
`and filled squares for cis. Also shown are the dihedral angles of the X-ray structure after minimization (X). The numbers of the dihedral angles on
`the abscissa correspond to the list in Table !\.
`
`0.1 kJ /(mol·A). After elimination of duplicates and high-energy
`structures, 21 distinct minima were found.
`The energy minimizations were performed using a dielectric
`constant of LO and a cutoff distance of 9.0 A for both van dcr
`Waals and electrostatic interactions. The calculations were
`performed with the BATCHMIN program, Version 2.7, on a DEC
`VuxStation 3500 minicomputer.27
`Results. The dihedral angles for the macrocyclic ring in the
`21 conformational minima found during the search are listed in
`Table II along with relative potential energies. A distribution of
`the dihedrnl angle5 is represented in Figure 3. Both Table Tl and
`figure 3 also contain data for the conformation from the X-ray
`
`(27) BATCH~HN is the noninteractive pan of the MACROMOOEL
`mo!~ular modeling program. Mohamadi. F.: Richards, N. G. J.; Guida. \V.
`C.: Lisk.amp, R.; Lipton, M.; Caufield, C.; Chang, G.: Hcndricl:son, T.; Still,
`W_ C. J. Comp. Chfm. 1990, / /, 440.
`
`structure after minimization. The 21 energy minima include both
`trans and cis amide isomers. Stereopictures of the lowest energy
`trans (1) and cis (2) forms are shown in Figure 4, along with the
`energy minimized X-ray structure; corresponding pictures and
`coordinates for all 22 structures are available in the Supplementary
`Material.
`It is emphasized that we did not attempt to locate all the
`conformational minima of FK506; to do so would have nece'>Sitated
`a much longer search as well as the inclusion of variations of the
`exocyclic and six-membered-ring torsions. However, the 21
`structures may be expected to be representative of the ]ow·energy
`conformations of FK506.
`All 21 structures arc reasonable in that no bonds or angles are
`unduly strained. Somewhat surprisingly, the X-ray crystal
`structure is not among these 21 structures. Its energy, or rather,
`the energy of the minimum closest to it, is 14.7 kcal/mo! above
`the energy of the lowest minimum found. A minimization was
`
`Roxane Labs., Inc.
`Exhibit 1016
`Page 004
`
`

`
`Comp111a1ional Studies on FK506
`
`J. An1. Che111. Soc., Vol. 113, A'o. 25, 1991 9487
`
`conformation
`IS
`II
`16
`14
`13
`12
`10
`3l
`167
`45
`18
`176
`19
`51
`-98
`-97
`-103
`-91
`-106
`-98
`-95
`-167
`-172
`5
`167
`2
`167
`180
`-110
`-136
`-120
`-172
`-80
`-90
`152
`61
`72
`84
`69
`15
`78
`69
`112
`115
`112
`173
`170
`174
`175
`-177
`-174
`-169
`-175
`-170
`-178
`-168
`-171
`-169
`66
`76
`72
`159
`66
`-43
`57
`79
`60
`173
`177
`60
`-162
`-178
`-141
`-137
`-105
`-166
`-170
`11
`61
`-57
`62
`84
`63
`81
`-104
`-84
`-105
`-127
`-I 13
`-3
`-98
`-177
`-179
`-175
`176
`178
`180
`178
`-91
`-112
`-!01
`-122
`-106
`-118
`-120
`-22
`-72
`52
`-57
`-64
`I 39
`68
`154
`-133
`-154
`-!69
`107
`115
`125
`-176
`-70
`70
`78
`173
`146
`-58
`-l 57
`-153
`-93
`-76
`46
`-68
`-7]
`-45
`-40
`-55
`-55
`-47
`-41
`43
`-154
`-158
`-151
`-\52
`-149
`- I 5 l
`69
`-!70
`177
`-178
`-174
`l7J
`178
`179
`27.4
`27.4
`27 .0
`26.8
`26.7
`26.4
`26.7
`dihcdr.i.l angles are not included in the search. ti Energies in kcal/mo!.
`
`17
`-172
`-99
`-2
`-79
`89
`176
`-169
`18
`123
`64
`72
`-108
`180
`-115
`-46
`-142
`-68
`-63
`-23
`-155
`-167
`27.5
`
`18
`173
`-99
`-2
`-115
`72
`176
`-l77
`86
`-57
`71
`171
`-18
`111
`-122
`I 05
`-142
`173
`-68
`-38
`-150
`177
`27.7
`
`19
`19
`-92
`171
`-126
`64
`170
`-173
`58
`54
`-170
`55
`-113
`174
`-134
`98
`-130
`177
`-61
`63
`65
`160
`28.2
`
`20
`89
`-98
`-3
`-88
`81
`111
`-174
`82
`65
`-177
`65
`-122
`-177
`-117
`127
`-147
`172
`-162
`51
`-155
`-165
`28.3
`
`21
`15
`-94
`167
`-177
`69
`173
`-177
`14
`62
`-174
`68
`-117
`-179
`-117
`101
`-108
`-64
`-91
`-45
`-155
`175
`28.3
`
`X-ray"
`124
`-93
`0
`-78
`87
`176
`-176
`80
`63
`111
`63
`-128
`-174
`-122
`I 35
`-132
`63
`-169
`180
`158
`173
`32.)
`
`also carried out starling with the geonlelry of bound FK506
`obtained from the X-ray crystal structure of its complex with
`FKBP.ll The resultant structure had an energy 21.5 kcal/mol
`above the lowest energy minimum.
`Interestingly. the cis isomers appear to favor a perpendicular
`orientation of the adjacent carbonyl groups (C8-C9 dihedral
`angle), while the trans isomers are more tolerant of an anti
`orientation in this position. Another noticeable difference is that
`the conformation about the Cl-C2 bond is always gauche-like
`in the trans isomers; it is usually anti in the cis. These trends,
`of course. only reflect the results for the 21 structures. However,
`these torsions are in the binding region of FK506, and the tend(cid:173)
`encies they show may be important in view of the observation that
`only the trans isomer is bound to FKBP. 11 No other trends are
`strikingly apparent which differentiate the cis and trans isomers.
`Not surprisingly, the four torsions excluded from the search
`(those around the C2-N7, Cl0-05, 05-C14, and C19-C20
`bonds) remain in their initial conformation (Figure 3). Three other
`torsions also show no tendency for variations, namely those around
`the C9-Cl0, C20-C21. and 01-Cl bonds. The anti orientation
`for the last torsion means that the ester is always in the Z form.
`/\iost of the rest of the torsions are clustered into easily identifiable
`gauche and anti conformations. Of course, the torsions flanking
`the double bond (Cl 8-Cl 9 and C20-C2 l) are expected to cluster
`around skew and syn conformations instead, and they do, except
`that C20-C21 is a torsion that remains skew in all 2! structures.
`The torsions flanking the isolated keto group (C21-C22 and
`C22-C23) show a rather large spread of \•alues, and do not appear
`to be easily categorized into gauche or anti conformations.
`A major reason for the absence of the X-ray crystal structure
`from the set is the presence of intramolecular hydrogen bonds.
`The actual X·ray structure does not have any intramolecular
`hydrogen bonds, but one (06-H J ... Q4) was formed upon energy
`minimization (Figure 4). This hydrogen bond is present in all
`the other structures; furthermore, !6 of these structures have
`additional hydrogen bonds involving OJO-H2 as a donor (Table
`Ill).
`In the actual crystal structure, 0!0-H2 acts as an inter·
`molecular hydrogen bond donor to 09ofa neighboring molecule. 7
`Jn addition. one water molecule was located in the crystal which
`forms the hydrogen bonds 06-H\ .. ·0{\V) and 0(\V)-H···04,
`preventing the formation of a direct hydrogen bond between
`06-H I and 04. The water molecule also forms a hydrogen bond
`to 03 of a neighboring FK506 molecule. Another intriguing
`difference in the crystal structure is the anti orientation about the
`C25-C26 bond. The orientation is invariably gauche in the 21
`
`Table ill. lmramolecular Hydrogen Bonds in the S1ructures Found
`in the ~fonte Carlo Se.arch"
`conformation
`1
`2
`J
`4
`S
`6
`7
`8
`9
`10
`1 I
`11
`13
`14
`JS
`16
`17
`18
`19
`20
`21
`X-ray
`(minimized)
`0 A h)drogen bond is deemed 10 exist if the distance between the
`hydrogen and the acceptor is !c;s than 2.5 A and the donor-hydrogen(cid:173)
`acceptor angle is greater than 1209
`
`hydrogen bonds
`06·-H\ .. ·04; OIO-H2 ... Q8
`06-ill· .. 04; OJO-H2 ... os; OIO-H2 .. ·0I
`06-Hl···04
`06-HJ· .. 04; OlO-H2 ... 09
`06-Hl .. ,04
`06-H\ ... 04; O\Q-H2· .. 05; OIO-H2 .. ·0I
`06-HJ· .. 04; OJO-H2· .. 09
`06-Hl .. ·04
`06-Hl· .. 04; O!O-H2· .. 09
`06-HJ .. ·04
`06-Hl·••04; 010-H2···09
`06-HJ· .. 04; 010-H2 .. ·08
`06-H\·•·04; OI0-112 ... 09
`06-Hl· .. 04; OJO-H2· .. 03; OIO-H2 .. ·01
`06-Hl· .. Q4; Ot0-H2···01
`06-Hl· .. 04: OIO-H2 ... Q9
`06-Hl· .. 04
`06-Hl···04: O!O-H2···09
`06-Hl· .. 04; OIO-H2· .. 09
`06-HJ· .. 04; OIO-H2 ... Q9
`06-HJ· .. 04; OlO-H2 ... 05
`06-H!· .. 04
`
`•
`
`olher structures, which facilitates the intramolecular hydrogen
`bonding with 010-H2.
`~lolecular D)'namics
`Procedure. In order to investigate the dynamical behavior and
`salvation of FK506, a molecular dynamics (l\1D) simulation of
`the molecule in water was performed. The energy-minimized
`X-ray structure was taken as the starting point for the MD sim·
`ulation. This structure was immersed in a box of TIP3P water
`molecules. 28 Any water molecule with its oxygen atom closer
`1han 1.5 A or with a hydrogen closer than 0.5 A to any FK506
`atom was removed. Any water molecule with its oxygen farther
`away than 7 .0 A from any solute atom in the x, y, or z directions
`was also removed. resulting in a system consisting of the solute
`
`(28) Jorgensen, \V. L.; Chandrasekhar. J.; Madura. J. D.: lmpey, R. \V.;
`Klein, M. L. J. Chem. Phys. 1983, 79, 926.
`
`Roxane Labs., Inc.
`Exhibit 1016
`Page 005
`
`

`
`9488 J. A111. Cht>m. Soc., Vol. 113. No. 25. 1991
`
`l'ronata and Jorgt'lt.ft'll
`
`Figurl' 4. Stcrcorcprcscntations of the !o'>I c.;t energy tr JllS (top) and cis (middle) isomers found in tl1c conformJlional ~car,·h. 'I he i.:Onforniation obtained
`frorn energy minimi1ation of tlH.' ~·r>~lal 'tructurc is ~hown at the bu!tom.
`
`and S58 water molecules (17.34 total atoms) in a box whose initial
`dimensions were 28.6 X 28.2 X 21 A A. From this ~iint on,
`periodic boundary conditions \\ere imposed on the. system.
`Equilibration of the i'.)'Stern was achieved in several phases. To
`begin, the water molecules were energy minimized for 100 steps
`using a ~tecpc~t tlc~ccnt algorithm, whik keeping the solulc gc(cid:173)
`oml!tr) fi)..cd. Thi! 5.olvcnl was then subjected to 5 ps of con~tant
`volume d)n.unic~. during whirh the tcmpcraturi: of th.: sy~tcrn
`wu~ rai.-,cd from !00 to 298 Kin a stepwise fashion. An additional
`4 ps of \imulation fullu\\Ctl. keeping the temperature :n 298 K,
`with the ~\llllll' ~till fi;...cd. Fi11:1lly, 1 P' 11f dynamic\ wa\ performed
`in whidi tho.: s<Jlut..: 1\a~ ;1llu11cd lo move.
`An \l[) ~imubti\lll was then pcrfnrmcd 11n thi\ c4uilibr•1tcd
`\y,tcm at constant temperature \298 K) and prl'ssun: (!bar=
`0.987 atm) for 200 p:;. Data anal;:~is 11.is cariicd out on the l.'ntirc
`1r;ijcc\Ory; then~ 11,1s no significain transient bch;wior during the
`e:irly p;1n or thl' simulation th.it might have resulted from in1-
`pcrk..:1 i.:4ui!ibralion
`
`All bond length\ and the 11-11 distance in water molccuks were
`constrained In their equilibrium value~. using the SllAKE al(cid:173)
`gorithm with a tokr.-ince of 0.0004 A/~ Nonbonded intcrnctions
`were c-,1!culatcd using a spherical n':Sidue-bascd cutoff, with fK506
`defined a~ a ~inglc residue. The cutoff distance was 9.0 A. To
`accdcratc the calculations, a nonbondcd pair lbt w;is used and
`u11dated every 10 steps. The time step u~cd in the simulation was
`2 fs, and data culk\:tion was r~·rformcd every 0.1 ps. The sim(cid:173)
`ulation wa\ jX'rfnrmed using 1\r-.·l