ࡱ> zaԑ]>]F( ww`!x&a/ } 0F1 8ڃ`ZPxՙ lUg̜c[Ԧix 6njBB AlX`E` "4 V@$TBTbDB JP+Y *M;sοjisfΝ;5 0hH, 12Si)Kws35ypfT2IJ6[27YOciWѕp~{&4ܔ|S>Y5k1t0޿_JB=t[^dO^jI,SrLH=SP2[am[4 rjjjWUmqxrP %)hgJ#K2rdkmKFz\uwnkw#\%wwWe2%H2WIq́ 7\ Hzz%. -;w n-. vsRpmpmpSH\@55 tJp.ÕGapC5grWe2zamuժBzLf,?ZV?sZL[@;@ZЧpྂk1g ;2*:Vo,ͶU1MP3m7O0M21yf)߼Tn:|}:mdڶM `ʲnazĺɺz^E43LQg.BZGLZL?9;h`ja>a1Og 8|]@LOQoJIHc>dȵ|5$/sSӫTTG˙6- $c`bVI7s7MnWW/q7mblSIm&1)VFlܤ'1 )C HCE\؅T|"1sldsfΝf̮`1?n;3 JK6s\VZBXE`sOGP=+,!ۗQgۮT 4q{nlC)l6?Oխդ`=vkkkwC/mHNT}[EhN\-  7$]^J}qRٰ;ǿ LĜIt3'B㘛wR㵆Av85 ~qٳu=TG㴆Ӛ83b~XQ&J>+$M)aOnǻO/' ѷMm]}um ڌ1gᲉ20ca^ m\&qUwߏgdO1CͯT7p KM먽aGXnů*2qڌ6|/bXtکj6|%X9~vS˭r&P c6j2T5%NxIq,,4zlBYdsC:q٩:rnC˘}lI.K eAKwb Sj38g1+`KR5G%a%66{O8pO!>=Br~PXϹnڅ3|ҾnD[s7 5[jzp}&Qr\g y>fq4B;&c0{j^2kRx#;#;q5g2yߏyɼ|xy'Q/ ]7Q'g+Vӌ1WXvR#5x0Ey{CUhx<~;Jy<, G51W4;yQSTG-ßƏ;㪉ϝ>o3d ;4dE{4ά:>[=d͊Vu 3}؁o!A~1{WSrov+p*vtKwo.mNZ+m[CQ}!sröa[r$؊$6:ëHwFu-|Nzؒ$c$--زփĖװ=ۢ 4l%l%ƶ,l=Hl'R5ب [6هL0(P8iTV3>N}PզM[V|䫮OJ c!X(mYPk8ca, -a)z c+*zX BiK(J=HlI [[ [[FÖalYaAbkmQzJc[$65BP6R6u}Ov z%w:߫3aҎψyOz2{h׼%v~t~Po p=\`! >:Zx 37hPE 8ڃ`Z x\kl9޻kB-@AF&@iU C # *AQä-u T, ARbj"?isvgC%;ߜsٙ4Ê͋y?QO[Lbh>x+'B·v+;#N: 4=hl'E: KK-|.˹ћ9@ze^Ƌ޽Sa_5y_D69Ur`<u:翇ŇP{gK47g=7NQQQr\ͣKvԺ>dX7Kgwφcf0^In#NP3֛^x}{u:耫QW 9/00!!w)!7%:\b*KĝX"W$K͊\bٺsft.+BZᴔuH2­}7;xgͶNc a1-rl; >-Y&Ewq^57'hc>R3or5^ح2e\NWoB. \q;]sZqN.WBOc:rƪ *Gp+cPMPe"nR`p:a+T]r5,$ 8TUsjgYj} hR,zvuXO3 *C:cE6V3 jǜ}0UUOKY7yRch aXo_tf%+f.TZE4!)n Bu5;c` +&B9c{Hj6co%>p9?|RwW ePA3mv6kUdYy͗:3*"nQ7XNPɊY(3Ԩ2$ FI}:rj4Êl,at睮C_ʊqp1X;38np*zVV>Fj?rj2A%5t5X$E+f1J&I'1_@Rs\U8]UX sXuUW+X$Eb\!Tj5A "mB]SNWxARd錗jnpj I1_2N.zW:]:܀0 jru#$!c,RjǐX(G=NW{Hj 0G]wӔVX."4~jOHM̧NW9MOE >q+~]Hm=~~u*VW/uo3VsU,QhGaWW?rq9g3J_MwzGCR1]U{*T8f2J͝t*["n TgH ް]3M"H(E'bg0]H1S`&ԯ2TW :=%u\tUjψ+j- n5j)fXXEjbrJd)R̃kH(5jޢHQƸ@ {8H [(r*RIgb tک)*p#sW)RRsA%[[b:($Lf*Rg27\\ ׹8Q`w| 5)"vi3VCuI7\5z(Iъ/1J] go˘Ier*fl&E(F >^5 3XQ'>q:/jV/R9];MP(uZ t:jRdQFR[i+Rܴ"#^sVk~cH zwl'8I7;:M8tufퟚcX&:38TYq3Qt횛 )V2F.zQt9`Hq]rUtUj]TR\Wk\$洭;R܄r#P 6Kq)oDHQPj:H;]Gzg"3Jͼaը&!^ȸ ntiRpw^5NW5*s B( NW j1Bc~-c/Uv %E cmbwp󢤘ͼ椹\I2H7}]ݫ̳w3~>=?=m)_Ȑd* fL0Cf\2 <2 fl O2,UD鳟Awx7s]~tN?1)wGRՊ)\&n\ w7(ug~q#^&ww>Eu,ҤnO&cH䢢g-|j =86=\$V{1Uq/V`+z/V'G[|YjDލ M3oN[;ŕ0kx@=HәyFepb7Th%F!xZOU?Bۄ~ Q,mjTۤ1mh(5PMӧ}E'#f7K!$6>Kf7?@w(Ɛ83s2;{3w9;gV=vSqW/s`e6ل~~;sz[@R{M$tbپI}%}KCM6'P'Updci> /rX,aBznH<٢X3ߋjJF_C_O_Q meYEZ ]0w֞Z;Vv~C?D~ktWG?&h>}7Š}<)G i7оy? dw%ݏNj߷4;{lͶ&wPqjq0u܆k;NI l<lDA LHLj;d<&Aqg2>p 0g cS0KNOqe\Ӟc|;4v}:|Nu1rVdI}::T3u5:D6ՙV,ϰ{YFxNVmjҭx7.,Ioԝ}DXrg+F+Ӕ6a-m߿ٶhضXƶE-Qڶnض^ƶu-QF56lm~nU|  mmm)öTRlۼ(meؖeۖGi[ް-_ƶ<۶!<>J{n[ ȶVհMcwcR UT<1Q5ȏ#H}( k4hm0Oz*uWc hVq|k|d \#ηo-olv6%-p׎8FDpUm[ȷηJ>;|kVeGom#Z"8*ׁm[K[%KQ#ȷϓ˅r|Ojf빳Wj:5Fgy~kjDӹǴK^?C5R}Q+?{Svv_c@=kL;_$/ %F!9xZMlUqqӨFr UDEvH8-ITFI(F g H $.= H pɖq*J"xy_w̳~o<3;{o-MpoQKk(^G췈c T=HH phm6H/d"T?:d?4BE.=ѕl.Pw?Bq(qS( vd n|/IՆjM)]ѳFmt\ko B K_R7\蜄E3?kQxߝƉp {G֍?!~E/Q=u-ǝ< q:]{oᛐx!#n݋ǿC˓{/rj|-JC?} wrk|"~e-gł6/%&}dsxOsl0;&t]ńx  C>v[_lF{(7m!HC,=U$]-v4AjFmH cs/lM?c,aclD96Y_ ylhmN ['\?e}"2uI13lDga'fW5Ӈ$a_L0>)ki,qQ} ~6,`3 !f?6K͊Ol 6>\0P%&͢l.{CYclr>Ycl.ج ?f"6kMU9 qhm+6یMm&lckOKSos]?mi62u199`(8jTyk=VA@p8~m Ŧd ەeyO;ȱbI50Yxn5+x)fHV4V*¾ ҷ[o9mK8|ѳ[om7iC6zfE¤@D9966LUdBlPǐY-l4)!72O'l;16RZpoMolkV)s- 8ZZ5wyZ[K˷ηR6H+|kiVgo--ߚ;J<ף[Sx[)zp|k |+e\ηoMo,F ~4\(nlvΝʹ[ѩڤ= 8$ҹxi!;_6>{Sog sB(o / 00DArialngs-,v0,0$DBrush Script MT,v0,0$B DTimes New Roman,v0,0$0DWingdingsRoman,v0,0$@ .  @n?" dd@  @@`` BL5    !"*#$ %&'( )*)+,-. /#0123 456789:;<=>?@`r$ԑ]>]Fi$$2$x&a/ } 0F1i"$W *Sw=2$>:Zx 37h% "$әyFepb7TP "$L;_$/ s] 0AA 3fff@fw$O ʚ;Nk7ʚ;g4JdJdD0ppp@ <4ddddph0\-<4!d!dph0\- 80___PPT10 ?  %7=HTS Triage Philosophy, Strategy and Tools 10 September 2007P> 0 $ Paul KirchhoffGetting Started$, I just officially started my position in the COP ~60% of my effort will be with the LSI/CCG Further development of Mscreen Implementation and development of new tools Leading HTS Triages Analysis of HTS data for follow up Preferably in an interactive meeting with all project members present!^b$I^b$,       Today s Talk $ Briefly describe my experience My background in HTS Triage Provide some basic observations from this experience Philosophy My approach to dealing with HTS data Strategy Triage tools Some examples of triage criteria Steps after HTS Triage ZZ7Z Z&Z ZZ"ZZ 7 &   "       Comp Chem Experience$  Doctoral work in computational chemistry Molecular mechanics simulations Internship at Agouron doing structure-based design Method development at Accelyrs (software company) Scoring function for protein-ligand binding Module for generating pharmacophore queriest*ZUZ3ZZZ*U3Z Lots of Experience with HTS Data!!$ Responsible for conducting triages on ~20 projects Triages on primary, confirmation, and IC50 data Contributed to the triages of many other projects Therapeutic areas Antibacterial  CNS Cardiovascular  Inflammation Size of the screening collection Initially ~800 thousand samples Last screen I triaged was 1.6 million samples4PdPP5P"PPP4) 9#  #"#P#Some HTS Philosophy$r HTS is systematic serendipity Throw a bunch of compounds at a target Hope some  stick Serendipity is the great strength of HTS Can provide hits which are active in unexpected ways(*6(*  5# Cream Floats But So Does Poop!$` In a primary screen, most samples should be inert We look for outliers Inhibitors or activators Some of the outliers will be poop No matter how  clean your screening collection or assay! HTS is a process with ample opportunities for errors Some quality sacrificed for quantity (~50% repeat rates)IPP#PPI#q # 9 #  HTS is a Significant Investment $ Don t get cheap at the last minute You can always go back and test more  sort of New batches of protein, cells, reagents, sources of compounds, equipment, etc. add inconsistency Remember HTS is a process and little changes add up Follow up on as many compounds at each stage In one batch as is practical Doing so at least helps normalize your resultsTZZ.ZNZL.0 #  A Successful Screen$ Is Not Getting the best confirmation rate (precision) possible Identifying the most potent compounds J. chem. Inf. Comput. Sci. 2001, 41,1308-1315 Identifying a drug right out of the box If our screening set was larger a" One more spin and I ll hit the jackpotZ`Z/Z)ZJZ`)#J #   A Simple Strategy for HTS Triage!!$ At each step of triage, follow up on as many compounds as is practical, deprioritizing the  obvious junk first*qp#  Triage Funnel$ Filtering on Potency$ #LFiltering on Potency  How They Did It''$ ' Correlation of %Effect and pIC50(!$ $ %VFiltering on Potency  How Should We Do It?,,$ &Filtering with Other Experimental Data''$$ Promiscuous samples/compounds  general Hits in multiple screens indicate A problem with the sample or compound Track no. of times a sample is a hit  PI s definition of a hit Currently less than 100 samples have hit in e" 4 screens May generate a list based on Std Dev)Z#Z'ZAZ_Z(##'#A#_ # !Calculated Properties$v Lipinski s  Rule of 5  actually 4 parameters Poor absorption or permeation are more likely when: No. H-bond donors > 5 Expressed as sum of OHs and NHs No. H-bond acceptors > 10 Expressed as sum of Os and Ns Molecular weight > 500 CLogP > 5 or MLogP > 4.15 (Moriguchi LogP calculation  not measured LogP)04!#C/4#!# #  # ##C#(JCalculated LogP  You re Asking a Lot&&$< LogP - log of the n-octanol / water partition coefficient A measure of hydrophobicity Larger, more positive the number the more hydrophobic LogP of +0.5 H" 3.3% solubility in pure water LogP > +0.5 insoluble in pure water Foye s Principles of Medicinal Chemistry, 5th edition, 2002;ZZeZ%Z=Z:#e#% # ,#+ ###*HCalculated LogP  It s a Guesstimate%%$ ,JCalculated LogD  You re Asking More!&&$ LogD - log of n-octanol / water distribution coefficient LogP assumes the solute is un-ionized! Should use LogD if your solute is ionizable! According to the Wikipedia (with intrepidation) 80% of drugs are ionizable:W19# '#'#### #  # .All Calculated Properties$ Require accurate, consistent representations of structures Easy to achieve with 10 s or 100 s of structures Very challenging when dealing with 100 s of thousands Often inconsistencies are seen when looking at extremes B<;#0Polar Surface Area$ 3D PSA 3D conformation and surface area calculated TPSA (Topological Polar Surface Area) A lot like LogP calculations Uses a list of fragments or atom types each contributing x2 of PSA Determine how many of each type is in the structure Add up the PSAZ-Z'ZZZ-#$ # #<#+M#2Structural Alerts$ Structural queries Used to Flag compounds containing problematic motifs Useful in the prioritization of screening hits Evaluation of compounds for purchase or synthesisJZZf3#4Structural Alerts$S Two sets are currently implemented in Mscreen Black Flags Filters used to exclude undesirable chemical matter from the NIH Molecular Libraries Small Molecule Repository (MLSMR) Red Flags Additional motifs that tend to be highly reactive or highly toxic Likely to cause in vitro problems May implement Yellow flags in vivo problems/P PxP PfPPP/ x C #  #  #  #  #  # 6Structural Alerts$ 8 Wrapping Up $ HTS Triage can be a fairly subjective Some accuracy and precision are sacrificed for quantity Choices are not always obviousD'Y'9 # Questions? $ /     " $ &)+-/13579}  ` 33` Sf3f` 33g` f` www3PP` ZXdbmo` \ғ3y`Ӣ` 3f3ff` 3f3FKf` hk]wwwfܹ` ff>>\`Y{ff` R>&- {p_/̴>?" dd@,|?" dd@   " @ ` n?" dd@   @@``PR    @ ` ` p>> ?(    3 ry @A#"    6  `}  T Click to edit Master title style! !  0  `  RClick to edit Master text styles Second level Third level Fourth level Fifth level!     S  0x ^ `  X*  0< ^   Z*  0 ^ `  Z*  H̟#" `N. A PDK Sept 2007 2 H  0޽h ? 3380___PPT10.P Default Design0 <*(  < < 0lz$ P  $ X*  < 0|~$   $ Z* d < c $ ?   $ < 0pM$  0 $ RClick to edit Master text styles Second level Third level Fourth level Fifth level!     S < 6($ OPn  $ X*  < 60$ O n $ Z* H < 0޽h#6 ? 3380___PPT10.@݌}  $(  r  S ,  r  S Ɛ @   H  0޽h ? 33___PPT10i.Pc+D=' = @B + (Z(  (x ( c $p$@ $  (  6$"`  $  ( 6X$ "` ,$ 0 Z It is NOT my job to dictate how data will be analyzed It is your project, your data You have the most to gain or lose by its success or failure My job is to provide you with the tools and organization to make the best decisions possibleN7\^7\^H ( 0޽h ? 33___PPT10.Pc+҇X*D.' = @B D' = @BA?%,( < +O%,( < +D ' =%(D' =%(Dp' =A@BBBB0B%()))D' =1:Bvisible*o3>+B#style.visibility<*(%(D' =-o6Bwipe(up)*<3<*(+8+0+($ +c zr H (  Hx H c $$@ $  H  6$"`0  $  H 6$ "`   > H H 0޽h ? 33___PPT10i.Pc+D=' = @B +  .&@P(  Px P c $`@ `  P  6`"` ` f P 6T` "` ,$ 0  Research Scientist for a major pharmaceutical company Characterization of large collections of  drug-like structures HTS Triage Analysis, organization and presentation of data to project teams Development of best practices and supporting software Hit-to-Lead Design of combinatorial libraries Development of best practices and supporting software Iterative screening7PMPyP PZPP7My ZH P 0޽h ? 33___PPT10b.Pc+m:D' = @B D' = @BA?%,( < +O%,( < +D' =%(D' =%(D@' =A@BBBB0B%(D' =1:Bvisible*o3>+B#style.visibility<*P%(D' =-o6Bwipe(up)*<3<*P+8+0+P` +  4,`X(  Xx X c $:`@ `  X  6;`"`  ` l X 6`=` "` `,$ 0 2 I ve seen a lot of HTS data There will be differences with the CCG data Targets  Sample Collection  Screening Technology Goals  Data ProcessingvJPPPJ +#  #H X 0޽h ? 33___PPT10b.Pc+_,D' = @B D' = @BA?%,( < +O%,( < +D' =%(D' =%(D@' =A@BBBB0B%(D' =1:Bvisible*o3>+B#style.visibility<*X%(D' =-o6Bwipe(up)*<3<*X+8+0+X` +:  XP`(  `x ` c $a`@ `  `  6db`"`@ `  ` 6 d` "`p,$ 0   Be careful not to undermine this strength By triaging on what an active compound  should look like Avoid biasing with models based on knowns or binding sites Pharmacophore models Applying models to identify problematic structures is fine Structural alerts+Pw<*#!v## ! ; # #H ` 0޽h ? 33___PPT10b.Pc+_,D' = @B D' = @BA?%,( < +O%,( < +D' =%(D' =%(D@' =A@BBBB0B%(D' =1:Bvisible*o3>+B#style.visibility<*`%(D' =-o6Bwipe(up)*<3<*`+8+0+`` + hL(  hx h c $`@ `  h  6|`"` `  h 6$` "` ,$ 0 D False Positives  labeled as active for wrong reason Random Don t repeat on retest  a hiccup in the process Mechanistic Will repeat over and over again Compound  reactive, toxic, colored, fluorescent, interferes with assay Sample  contaminated, wrong compound, wrong concentration6PP2P PP62   #@ h 6` "`,$ 0 P False Negative  missed active compound,)P(#H h 0޽h ? 33___PPT10.Pc+ 7D' = @B D' = @BA?%,( < +O%,( < +D' =%(D' =%(DD' =A@BBBB0B%(D' =1:Bvisible*o3>+B#style.visibility<*h%(D' =-s6Bwipe(left)*<3<*hD' =%(D' =%(DD' =A@BBBB0B%(D' =1:Bvisible*o3>+B#style.visibility<*h%(D' =-s6Bwipe(down)*<3<*h+p+0+h` ++0+h` + pP(  px p c $P`@ `  p  6$`"`  ` H p 0޽h ? 33___PPT10i.Pc+D=' = @B +  rjx(  xx x c $`@ `  x  6`"` ` D x 6X` "`5,$ 0  Is Identifying leads Compounds which are active for the right reason Good ADMET properties FTO Minimally identifying compounds which teach you something about your system Help get a crystal structure for examplePPMPMP*PMM#* # ^ x 6H` "`eP,$ 0 ^ Most compound series perish in drug discovery Ideally a screen will identify multiple series</P0P/0#H x 0޽h ? 33___PPT10.Pc+3ZjD' = @B D' = @BA?%,( < +O%,( < +D' =%(D' =%(DD' =A@BBBB0B%(D' =1:Bvisible*o3>+B#style.visibility<*x%(D' =-s6Bwipe(left)*<3<*xD' =%(D' =%(DD' =A@BBBB0B%(D' =1:Bvisible*o3>+B#style.visibility<*x%(D' =-s6Bwipe(down)*<3<*x+p+0+x` ++0+x` + 6(  x  c $`@ `    6`"` ` !  6` "`` ,$ 0 y Obvious is often very unobvious If you are familiar with US patent law Your invention has to include a change which is unobvious to a technically competent person in the field Job security for a lot of attorneys!P(PjP%P (j%#`  6` "`  ,$ 0 b There are tools which can help identify the obvious junk Compounds less likely to lead to a drug::P)P:)M  6` "` ,$ 0 O How aggressively these tools are applied depends on How hit rich vs. hit poor:5PP5H  0޽h ? 33  ___PPT10 .Pc+<'D ' = @B D ' = @BA?%,( < +O%,( < +D' =%(D' =%(DD' =A@BBBB0B%(D' =1:Bvisible*o3>+B#style.visibility<*%(D' =-s6Bwipe(left)*<3<*D' =%(D' =%(DD' =A@BBBB0B%(D' =1:Bvisible*o3>+B#style.visibility<*%(D' =-s6Bwipe(left)*<3<*D' =%(D' =%(DD' =A@BBBB0B%(D' =1:Bvisible*o3>+B#style.visibility<*%(D' =-s6Bwipe(down)*<3<*++0+\ ++0+\ ++0+\ +jE    )0 ~(  x  c $H\@ \ 1  6\ "` ,$ 0  IC50 Curve Fit Counter ScreensN P #B  6\ "`,$ 0 L %Effect Other Experiments Calculated Properties Id False Positives/Sampling2MP0l      ,$D 0@      fB  6Dp  `B   0Dp  @      fB B 6Dp  `B   0Dp    6\#\ "`,$ 0 ^Prioritization Development(P#l p    p ,$D  0ZB  s *D0 0 ZB  s *DP P ` ZB  s *D ZB  s *D nr  0"` 0  `  0p nr  0"` ` @ z  <("`( j   <8*\ "` ,$ 0 UPrimary2P'  <8/\ "`` ,$ 0 RDose2P' ! 63\ "`0 ,$ 0 X 100 s2P' " 67\ "`p,$ 0 R100k2P' # 6T<\ "` P ,$  0 V10 s2P' $ 6@\ "`` P ,$  0 l IDP FP'@l `   '  ,$D  0 % 6E\ "`P   r$Mscreen, SARNavigator, ZINC, PubChem2%P$' & 6tJ\ "``   R4P' ( 6N\ "`p0,$ 0 p MCCSLFP' ) 6S\ "` ,$  0 DSeries ExpansionPH  0޽h ? 33//___PPT10..Pc+"JDT,' = @B D,' = @BA?%,( < +O%,( < +D' =%(D' =%(D3' =4@BBBB%(D' =1:Bvisible*o3>+B#style.visibility<*%(D' =-o6Bwipe(up)*<3<*D' =%(D' =%(DF' =A@BBBB0B%(D' =1:Bvisible*o3>+B#style.visibility<*"%(D' =-u6Bwipe(right)*<3<*"D' =%(D' =%(D@' =A@BBBB0B%(D' =1:Bvisible*o3>+B#style.visibility<*%(D' =-o6Bwipe(up)*<3<*D$' =%(D' =%(Dt' =A@BBBB0B%()))D' =1:Bvisible*o3>+B#style.visibility<*%(D' =-s6Bwipe(down)*<3<*D' =%(D' =%(D@' =A@BBBB0B%(D' =1:Bvisible*o3>+B#style.visibility<*!%(D' =-o6Bwipe(up)*<3<*!D' =%(D' =%(D@' =A@BBBB0B%(D' =1:Bvisible*o3>+B#style.visibility<*%(D' =-o6Bwipe(up)*<3<*D' =%(D' =%(DD' =A@BBBB0B%(D' =1:Bvisible*o3>+B#style.visibility<* %(D' =-s6Bwipe(down)*<3<* D' =%(D' =%(D@' =A@BBBB0B%(D' =1:Bvisible*o3>+B#style.visibility<*#%(D' =-o6Bwipe(up)*<3<*#D' =%(D+' =%(D' =A@BBB B0B%(D' =1:Bvisible*o3>+B#style.visibility<*$%(D' =+4 8?dCB0-#ppt_w/2BCB#ppt_xB*Y3>B ppt_x<*$D' =+4 8?dCB0-#ppt_h/2BCB#ppt_yB*Y3>B ppt_y<*$D' =%(D' =%(DD' =A@BBBB0B%(D' =1:Bvisible*o3>+B#style.visibility<*)%(D' =-s6Bwipe(down)*<3<*)D' =%(D' =%(D3' =4@BBBB%(D' =1:Bvisible*o3>+B#style.visibility<*%(D' =-o6Bwipe(up)*<3<*D' =%(D' =%(D9' =4@BBBB%(D' =1:Bvisible*o3>+B#style.visibility<*'%(D' =-u6Bwipe(right)*<3<*'D' =%(D' =%(DD' =A@BBBB0B%(D' =1:Bvisible*o3>+B#style.visibility<*%(D' =-s6Bwipe(down)*<3<*D' =%(D' =%(DF' =A@BBBB0B%(D' =1:Bvisible*o3>+B#style.visibility<*(%(D' =-u6Bwipe(right)*<3<*(+h+0+\ ++0+\ ++0+\ ++0+\ ++0+ \ ++0+!\ ++0+"\ ++0+#\ ++0+$\ ++0+(\ ++0+)\ +  F>@ (  x  c $0~\@ \   6x\ "` ,$ 0  Several ways a cutoff is chosen Pick a number Use 3 or 4 standard deviations as a cutoff More sophisticated statistics Use %effect and similarity tools to pick active compounds:"PP"  6\ "`p :4 %Effect  active vs. inactive cutoff Most important triage criteria Largest discriminating criteria Generally first criteria to be applied Not all important triage criteria J. chem. Inf. Comput. Sci. 2001, 41,1308-1315 Just testing the X most potent compounds not a good idea&Z ZIZ#ZiZ% I#        G  #  BA?"` ,$D 0H  0޽h ? 33w___PPT10W.Pc+-D' = @B D' = @BA?%,( < +O%,( < +D' =%(D' =%(D9' =4@BBBB%(D' =1:Bvisible*o3>+B#style.visibility<* %(D' =-u6Bwipe(right)*<3<* D' =%(D' =%(DD' =A@BBBB0B%(D' =1:Bvisible*o3>+B#style.visibility<*%(D' =-s6Bwipe(down)*<3<*+8+0+\ +b.  yq( (  x  c $d\@ \   HA?"`i,$D 0  6j "`@ ,$ 0 Samples Tested = 58588 Median = -2 Std Dev = 17 Cutoffs 3  = 3x17  2 = 49 %effect 1481 Hits 2.5 %yP!# '  '  '  ' ,?8l nht &hnt,$D 0`B  0D8cyhyt`B   0Dn``   <` j "`` E 1481 Hits P 'l ht 'ht,$D 0`B   0D8cht`B   0D   <j "` D916 Hits P '<  6j "`@ ,$ 0 DConfirmation Rates 3  H" 4  H" 5 #P#'#    '# ' Jl `   `  ,$D  0@ g   g  fB  6D8c\  fB B 6D8c y ZB  s *D|yg y   68!j "``   Lmore likely activeP'  6(#j "`< ` b  Lless likely activeP'3l p`  %p` ,$D 0ZB  s *Dpp`  ! 6p'j "``  R751 2P'ZB " s *Dp0 `  # 6,j "``   R286 2P'( $ 6(0j "`0`  (Control 3  = 42%BP '  ( 65j "` @ O ,$ 0 x4  = 4x17  2 = 66 %effect 916 Hits 1.6 %F=P'' ',9H  0޽h ? 33___PPT10i.Pc+E7D' = @B DP' = @BA?%,( < +O%,( < +D' =%(D' =%(D3' =4@BBBB%(D' =1:Bvisible*o3>+B#style.visibility<*%(D' =-o6Bwipe(up)*<3<*D' =%(D' =%(D@' =A@BBBB0B%(D' =1:Bvisible*o3>+B#style.visibility<*%(D' =-o6Bwipe(up)*<3<*D' =%(D' =%(D7' =4@BBBB%(D' =1:Bvisible*o3>+B#style.visibility<*&%(D' =-s6Bwipe(left)*<3<*&D' =%(D' =%(DD' =A@BBBB0B%(D' =1:Bvisible*o3>+B#style.visibility<*(%(D' =-s6Bwipe(left)*<3<*(D' =%(D' =%(D7' =4@BBBB%(D' =1:Bvisible*o3>+B#style.visibility<*'%(D' =-s6Bwipe(left)*<3<*'D' =%(D' =%(D3' =4@BBBB%(D' =1:Bvisible*o3>+B#style.visibility<*%%(D' =-o6Bwipe(up)*<3<*%D' =%(D' =%(D@' =A@BBBB0B%(D' =1:Bvisible*o3>+B#style.visibility<*%(D' =-o6Bwipe(up)*<3<*D' =%(D' =%(D7' =4@BBBB%(D' =1:Bvisible*o3>+B#style.visibility<*%(D' =-s6Bwipe(down)*<3<*++0+j ++0+j ++0+(j +  (  x  c $+B#style.visibility<*%(D' =-o6Bwipe(up)*<3<*D' =%(D' =%(DD' =A@BBBB0B%(D' =1:Bvisible*o3>+B#style.visibility<*%(D' =-s6Bwipe(down)*<3<*+8+0+j + )(  x  c $؇j@ j   6|j "`^ ,$ 0 ! For the time being suggest continue to use Mscreen to visualize plate data for systematic errors Std Dev cutoff on a per plate basis Perhaps also apply a minimum %effectX,P\P&P,\&0  6j "`  Ideally we use a smoothing tool which Accurately corrects for Any systematic error  for example row or column differences Signal drift in a run  assay Signal drift from one day to the next  entire screen Accurately, reproducibly, and objectively identifies hits'ZZZ;Z&;h  6j "` ,$ 0 j Do some retrospective analyses on existing data Determine how different choices would influence outcomes:1P:P1:H  0޽h ? 33___PPT10.Pc+RXD' = @B D' = @BA?%,( < +O%,( < +D' =%(D' =%(DD' =A@BBBB0B%(D' =1:Bvisible*o3>+B#style.visibility<*%(D' =-s6Bwipe(left)*<3<*D' =%(D' =%(DD' =A@BBBB0B%(D' =1:Bvisible*o3>+B#style.visibility<*%(D' =-s6Bwipe(down)*<3<*+p+0+j ++0+j +  `z(  x  c $@j@ j    6j"` j   6j "`^ ,$ 0  Promiscuous samples/compounds  specific assay tech Same assay technology  different target:5P*P5*`  6j "` T,$ 0 b Screens run explicitly to identify problem samples Mechanistic false positives  toxic, colored, fluorescent, etc. ADMET props  solubility, permeability, metabolic stability:4P~P4~H  0޽h ? 33___PPT10.Pc+PD' = @B D' = @BA?%,( < +O%,( < +D' =%(D' =%(DD' =A@BBBB0B%(D' =1:Bvisible*o3>+B#style.visibility<*%(D' =-s6Bwipe(left)*<3<*D' =%(D' =%(DD' =A@BBBB0B%(D' =1:Bvisible*o3>+B#style.visibility<*%(D' =-s6Bwipe(down)*<3<*+p+0+j ++0+j +*   (  x  c $j@ j    6j"` <$ 0 j z  6Pj "` P,$  0 Z According to Lipinski  Compound classes that are substrates for biological transporters are exceptions to the rule Orally active therapeutic classes outside the  Rule of 5 are: " Antibiotics " Antifungals " Vitamins " Cardiac Glycosides Advanced Drug Delivery Reviews 1997, 23, 3-25\P`P@PIP.P#^#?    ## # $$(#(,,,,000000 003  6j "`` ,$  0 GIf 2 or more parameters exceeded, structure labeled as possible problem(HPGDl      ,$D  0`B B 0DjJ  `B  B 0DjJ  H  0޽h ? 33!!___PPT10!.Pc+Ė;D ' = @B Dq ' = @BA?%,( < +O%,( < +D0' =%(D' =%(D@' =A@BBBB0B%(D' =1:Bvisible*o3>+B#style.visibility<*0%(D' =-o6Bwipe(up)*<3<*0D@' =A@BBBB0B%(D' =1:Bvisible*o3>+B#style.visibility<*0d%(D' =-o6Bwipe(up)*<3<*0dD@' =A@BBBB0B%(D' =1:Bvisible*o3>+B#style.visibility<*d{%(D' =-o6Bwipe(up)*<3<*d{D@' =A@BBBB0B%(D' =1:Bvisible*o3>+B#style.visibility<*{%(D' =-o6Bwipe(up)*<3<*{D@' =A@BBBB0B%(D' =1:Bvisible*o3>+B#style.visibility<*%(D' =-o6Bwipe(up)*<3<*D@' =A@BBBB0B%(D' =1:Bvisible*o3>+B#style.visibility<*%(D' =-o6Bwipe(up)*<3<*D@' =A@BBBB0B%(D' =1:Bvisible*o3>+B#style.visibility<*%(D' =-o6Bwipe(up)*<3<*D@' =A@BBBB0B%(D' =1:Bvisible*o3>+B#style.visibility<*%(D' =-o6Bwipe(up)*<3<*D@' =A@BBBB0B%(D' =1:Bvisible*o3>+B#style.visibility<*<%(D' =-o6Bwipe(up)*<3<*<D' =%(D' =%(DF' =A@BBBB0B%(D' =1:Bvisible*o3>+B#style.visibility<*%(D' =-u6Bwipe(right)*<3<*D' =%(D' =%(DD' =A@BBBB0B%(D' =1:Bvisible*o3>+B#style.visibility<*%(D' =-s6Bwipe(down)*<3<*Dv' =%(D' =%(D' =4@BBBB%(D' =1:Bvisible*o3>+B#style.visibility<* %(D' =+4 8?dCB1+#ppt_w/2BCB#ppt_xB*Y3>B ppt_x<* D' =+4 8?dCB0-#ppt_h/2BCB#ppt_yB*Y3>B ppt_y<* ++0+j ++0+ ++0+ + }u" (  x  c $l@     6@"`    6 "`` ,$ 0 i Generally two approaches to estimating LogP Sum up hydrophobic & hydrophilic contributions of Functional groups or atom types Empirically fit values No explicit treatment on conformation differences or ensemblest-P4PyP-4d,l P0  "P0 ,$D 0`  0P0 n  0"`P   0   7water 2   0p   ; n-octanol 2 n   0"`pn   0"`   n   0"`p n   0"`0 ` n  0"`P0 n  0"`0n  0"`P  n  0"`0  n  0"`0 p n  0"`p0 n  0"` @ n   0"`p  h ! 6) "`p ,$ 0  LogP = log {[solute]n-octanol/[solute]water} The solute is un-ionized! LogP = log [9/3] = 0.48 water soluble.ZEZ# + #+## '##H  0޽h ? 33  ___PPT10 .Pc+!)D= ' = @B D' = @BA?%,( < +O%,( < +D3' =%(D' =%(DD' =A@BBBB0B%(D' =1:Bvisible*o3>+B#style.visibility<*!%(D' =-s6Bwipe(left)*<3<*!D7' =4@BBBB%(D' =1:Bvisible*o3>+B#style.visibility<*"%(D' =-s6Bwipe(left)*<3<*"D' =%(D' =%(DD' =A@BBBB0B%(D' =1:Bvisible*o3>+B#style.visibility<*%(D' =-s6Bwipe(down)*<3<*+p+0+ ++0+! +^ xp(  x  c $K@  >  6M "` ,$ 0  Using calculated LogP values for triage Remember they are guesstimates! Don t automatically throw out all structures with 0 > LogP > 5 Look at the extremes Like to have an experienced medicinal chemist present in all triages)P!P@PPFP)!@:      F P0   `  0P0 n  0"`P   0L\   7water 2   0]  ; n-octanol 2 n   0"`pn   0"`   n   0"`p n   0"`0 ` n  0"`P0 n  0"`0n  0"`P  n  0"`0  n  0"`0 p n  0"`p0 n  0"` @ n  0"`p    6e "`  Different LogP calculators Can give very different results Expect to see differences From different algorithms ALogP, CLogP, MLogP, MiLogP, SLogP, XLogP Same algorithm implemented by different personsZ<ZZ+Z1Z#<##+#1#H  0޽h ? 33___PPT10f.Pc+n-`D' = @B D' = @BA?%,( < +O%,( < +D' =%(D' =%(DD' =A@BBBB0B%(D' =1:Bvisible*o3>+B#style.visibility<*%(D' =-s6Bwipe(down)*<3<*+8+0+ + #  0(  x  c $hz@     6<{"`    6` "`  ,$ 0 . LogDacids = logP + log[1/(1 + 10pH  pKa)] LogDbases = logP + log[1/(1 + 10pKa  pH)] J. Med. Chem. 1977, 20, 53-58$,PP,PPP       #l    ,$D 0  6 "`  BP LogD = log {[solute]n-octanol / ([solute]water + [solute]water )}QZ# # ++## # +#+# +#I  6 "`K +0  +ionized un-ionized,Z/++#+ /' l 0  ,$D 0f  60 t  6"`0 `   68` G  7water 2   6h   ; n-octanol 2 t   6"``P  t   6"` `P t   6"` p t   6"` @0 t  6"` 0 t  6"`  t  6"`0p  t  6"` p0 t  6"` P0 t  6"`P  t  6"`P  t  6"`p @nB  0"`P  nB  0"` @nB  0"` p nB  0"` @nB  0"`    6 "` P,$ 0 B LogD requires both a logP and a pKa value 1 guesstimate + 1 guesstimate probably `" greater accuracy\+PP:P+ '#H  0޽h ? 33  ___PPT10z .Pc+=J D ' = @B D ' = @BA?%,( < +O%,( < +D' =%(D' =%(D9' =4@BBBB%(D' =1:Bvisible*o3>+B#style.visibility<*%(D' =-u6Bwipe(right)*<3<*D' =%(D' =%(D7' =4@BBBB%(D' =1:Bvisible*o3>+B#style.visibility<*%(D' =-s6Bwipe(left)*<3<*D' =%(D' =%(DD' =A@BBBB0B%(D' =1:Bvisible*o3>+B#style.visibility<*%(D' =-s6Bwipe(left)*<3<*D' =%(D' =%(DD' =A@BBBB0B%(D' =1:Bvisible*o3>+B#style.vi