28
PI3Kinases: An Example of HoloWaterMap Modelling Daniel D. Robinson

PI3$Kinases:,An,Example,of,Holo$WaterMap, Modelling,content.schrodinger.com/Training+Material/WM/PI3K/PI3-Kinases_fi… · Perplexing,SAR:,Stereochemistry,Controls,AcKvity,and, SelecKvity

  • Upload
    others

  • View
    4

  • Download
    0

Embed Size (px)

Citation preview

Page 1: PI3$Kinases:,An,Example,of,Holo$WaterMap, Modelling,content.schrodinger.com/Training+Material/WM/PI3K/PI3-Kinases_fi… · Perplexing,SAR:,Stereochemistry,Controls,AcKvity,and, SelecKvity

PI3-­‐Kinases:  An  Example  of  Holo-­‐WaterMap  Modelling  

Daniel  D.  Robinson  

Page 2: PI3$Kinases:,An,Example,of,Holo$WaterMap, Modelling,content.schrodinger.com/Training+Material/WM/PI3K/PI3-Kinases_fi… · Perplexing,SAR:,Stereochemistry,Controls,AcKvity,and, SelecKvity

INTRODUCTION  –  PERPLEXING  SAR  AND  STRUCTURAL  DATA  

Page 3: PI3$Kinases:,An,Example,of,Holo$WaterMap, Modelling,content.schrodinger.com/Training+Material/WM/PI3K/PI3-Kinases_fi… · Perplexing,SAR:,Stereochemistry,Controls,AcKvity,and, SelecKvity

Perplexing  SAR:  Stereochemistry  Controls  AcKvity  and  SelecKvity  

Compound  ID   R1   R2   PI3K-­‐β  (nM)   PI3K-­‐δ  (nM)   SelecKvity  

(Fold)  

A   H   H   4   28   7  

B(U)   H   Me   23   468   20  

B(D)   Me   H   6   6   1  

C   Me   Me   102   580   6  

D(U)   H   Et   135   1233   10  

D(D)   Et   H   10   11   1  

E(U)   H   cPr   381   1610   4  

E(D)   cPr   H   2   2   1  

F(U)   H   iPr   615   1231   2  

F(D)   iPr   H   4   1   0.25  

G(U)   H   Ph   470   1984   4  

G(D)   Ph   H   4   5   1  

The  SAR  shows  two  clear  trends:    •  The  ‘up’  (U)  stereochemistry  

is  consistently  weaker  on  both  β  and  δ  isoforms    

•  The  (U)  stereochemistry  shows  moderate  selecTvity  for  PI3K-­‐β,  at  least  for  the  smaller  R-­‐groups  

Page 4: PI3$Kinases:,An,Example,of,Holo$WaterMap, Modelling,content.schrodinger.com/Training+Material/WM/PI3K/PI3-Kinases_fi… · Perplexing,SAR:,Stereochemistry,Controls,AcKvity,and, SelecKvity

Structural  Data•  Crystal-­‐structures  of  B(U)  bound  to  PI3K-­‐β/δ  were  available  –  These  structures  showed  that  the  criTcal  subsTtuents  made  few,  if  any,  direct  interacTons  with  their  respecTve  host  proteins  

•  If  a  direct  interacTon  could  not  explain  the  SAR,  maybe  a  solvent  effect  could?...  

Prepared  PI3K-­‐β/δ  crystal  structures    

ENTRIES:  PREPARED  CRYSTAL  STRUCTURES  VIEW:  DEFAULT  VIEW  

Page 5: PI3$Kinases:,An,Example,of,Holo$WaterMap, Modelling,content.schrodinger.com/Training+Material/WM/PI3K/PI3-Kinases_fi… · Perplexing,SAR:,Stereochemistry,Controls,AcKvity,and, SelecKvity

HOLO-­‐WATERMAP  ANALYSIS  

Page 6: PI3$Kinases:,An,Example,of,Holo$WaterMap, Modelling,content.schrodinger.com/Training+Material/WM/PI3K/PI3-Kinases_fi… · Perplexing,SAR:,Stereochemistry,Controls,AcKvity,and, SelecKvity

Why  Holo-­‐WaterMap  Analysis?  •  Classical  apo-­‐WaterMaps  concern  themselves  with  the  ligand-­‐induced  desolvaTon  of  the  host  pocket  – As  such  it  considers  the  water-­‐molecules  that  are  no  longer  present  in  the  pocket    

•  Holo-­‐WaterMap  analysis  does  the  opposite  –  It  focuses  on  the  effect  of  the  ligand  on  the  residual  solvaTon  – Depending  on  the  ligand  and  the  binding-­‐site,  the  ligand  may  stabilise  or  destabilise  the  residual  water-­‐molecules  •  Forming  a  bridging  interacTon  may  well  stabilise  a  water-­‐molecule  •  Trapping  a  water-­‐molecule  between  the  ligand  and  the  protein  is  likely  to  lead  to  significant  destabilisaTon  

Page 7: PI3$Kinases:,An,Example,of,Holo$WaterMap, Modelling,content.schrodinger.com/Training+Material/WM/PI3K/PI3-Kinases_fi… · Perplexing,SAR:,Stereochemistry,Controls,AcKvity,and, SelecKvity

Aside:  GeneraKng  Holo-­‐WaterMaps  –  From  the  GUI  

•  Holo-­‐WaterMaps  can  be  generated  immediately  from  the  WaterMap  GUI  panels  –  Place  the  structure  of  interest  into  the  Workspace  

–  Bring  up  the:    WaterMap-­‐>Perform  Calculations  Panel  

–  ‘Pick’  the  ligand  to  instruct  WaterMap  where  to  concentrate  its  efforts  

–  Select  the  ‘Retain  ligand’  opTon  –  Deselect  the  ‘Truncate  protein’  opTon  

Page 8: PI3$Kinases:,An,Example,of,Holo$WaterMap, Modelling,content.schrodinger.com/Training+Material/WM/PI3K/PI3-Kinases_fi… · Perplexing,SAR:,Stereochemistry,Controls,AcKvity,and, SelecKvity

Aside:  GeneraKng  Holo-­‐WaterMaps  –  From  the  Command  Line  

•  The  calculaTon  of  holo-­‐WaterMaps  is  beder  performed  from  the  command  line.  –  The  uTlity  $SCHRODINGER/utilities/create_wm_job  makes  this  quite  easy  

•  Firstly,  this  uTlity  allows  us  to  script  the  creaTon  of  many  holo-­‐WaterMap  jobs  – Useful  when  there  are  a  number  of  ligands  

•  Secondly  this  uTlity  gives  us  access  to  the  –extended_gcmc  opTon  –  This  opTon  more  exhausTvely  explores  the  solvaTon  surrounding  the  ligand  •  This  is  parTcularly  useful  for  holo-­‐WaterMaps  as  the  presence  of  the  ligand  frequently  generates  regions  where  it  is  difficult  to  equilibrate  bulk  water  

Page 9: PI3$Kinases:,An,Example,of,Holo$WaterMap, Modelling,content.schrodinger.com/Training+Material/WM/PI3K/PI3-Kinases_fi… · Perplexing,SAR:,Stereochemistry,Controls,AcKvity,and, SelecKvity

Holo-­‐WaterMap:  PI3K-­‐β  A  •  The  holo-­‐WaterMap  of  PI3K-­‐β  complexed  to  ligand  A  reveals  a  very  interesTng  looking  water-­‐molecule  –  This  water  is  ‘trapped’  between  the  ligand  and  the  ‘base’  of  the  acTve-­‐site  

•  The  posiTon  of  the  water-­‐molecule  is  such  that  it  would  be  influenced  by  the  R1/R2  subsTtuents  of  the  ligands  

Page 10: PI3$Kinases:,An,Example,of,Holo$WaterMap, Modelling,content.schrodinger.com/Training+Material/WM/PI3K/PI3-Kinases_fi… · Perplexing,SAR:,Stereochemistry,Controls,AcKvity,and, SelecKvity

Holo-­‐WaterMap:  PI3K-­‐β  B(U)  +  D(U)  

Holo-­‐WaterMap  PI3K-­‐β B(U)    

The  same  water-­‐molecule  is  found  in  the  complexes  of  the  other  ligands.    

Holo-­‐WaterMap  PI3K-­‐β D(U)    

The  general  trend  is  that  the  water-­‐molecule’s  energy  increases  with  the  R-­‐group’s  size.  

Page 11: PI3$Kinases:,An,Example,of,Holo$WaterMap, Modelling,content.schrodinger.com/Training+Material/WM/PI3K/PI3-Kinases_fi… · Perplexing,SAR:,Stereochemistry,Controls,AcKvity,and, SelecKvity

Holo-­‐WaterMap:  PI3K-­‐δ  A  •  The  same  water-­‐molecule  is  found  in  the  complex  of  PI3K-­‐δ  with  ligand  A  

Page 12: PI3$Kinases:,An,Example,of,Holo$WaterMap, Modelling,content.schrodinger.com/Training+Material/WM/PI3K/PI3-Kinases_fi… · Perplexing,SAR:,Stereochemistry,Controls,AcKvity,and, SelecKvity

Holo-­‐WaterMap:  PI3K-­‐δ  B(U)  +  D(U)  

Holo-­‐WaterMap  PI3K-­‐δ B(U)    

We  see  the  same  trend  of  increasing  water-­‐energy  with  R-­‐group  size.    

Holo-­‐WaterMap  PI3K-­‐δ D(U)    

However,  PI3K-­‐δ  appears  to  be  a  bit  more  sensiTve  than  PI3K-­‐β,  with  greater  instability  

for  a  given  R-­‐group.  

Page 13: PI3$Kinases:,An,Example,of,Holo$WaterMap, Modelling,content.schrodinger.com/Training+Material/WM/PI3K/PI3-Kinases_fi… · Perplexing,SAR:,Stereochemistry,Controls,AcKvity,and, SelecKvity

Intermediate  Conclusions  •  The  greater  acTvity  of  the  X(D)  compounds  vs.  the  X(U)  compounds  appears  to  be  explained  by  a  single  water-­‐molecule  –  The  X(D)  compounds  all  displace  this  water-­‐molecule  –  The  X(U)  compounds  acTvely  trap  the  molecule  between  themselves  and  the  ‘base’  of  the  acTve-­‐site  

•  Increasing  the  size  of  the  subsTtuent  of  the  X(U)  compounds  traps  the  water-­‐molecule  even  more  –  This  increases  the  energy  of  the  water-­‐molecule,  broadly  accounTng  for  the  loss  in  potency    

–  The  trapping  effect  appears  to  be  more  significant  in  PI3K-­‐δ  than  β,  this  may  account  for  the  observed  selecTvity  

Page 14: PI3$Kinases:,An,Example,of,Holo$WaterMap, Modelling,content.schrodinger.com/Training+Material/WM/PI3K/PI3-Kinases_fi… · Perplexing,SAR:,Stereochemistry,Controls,AcKvity,and, SelecKvity

SEMI-­‐QUANTITATIVE  HOLO-­‐WATERMAP  ANALYSIS  

Page 15: PI3$Kinases:,An,Example,of,Holo$WaterMap, Modelling,content.schrodinger.com/Training+Material/WM/PI3K/PI3-Kinases_fi… · Perplexing,SAR:,Stereochemistry,Controls,AcKvity,and, SelecKvity

Aside:  The  Origins  of  Holo-­‐WaterMap  Analysis  The  first  study  to  use  holo-­‐WaterMaps  for  ligand  scoring  was  reported  by  Snyder  et  al.*  They  studied  a  series  of  ligands  bound  to  CA-­‐II  and  correlated  the  

WaterMap-­‐results  with  ITC  

CA-­‐II  is  very  rigid  and  the  hydraTon-­‐paderns  around  the  ligand  were  highly-­‐conserved.  In  this  case  it  

was  easy  to  idenTfy  corresponding  hydraTon-­‐sites  in  pairs  of  holo-­‐WaterMaps  and  sum  the  changes  

in  water-­‐energies  

*Snyder  et  al.  PNAS,  2011,  108  (44),  17889–17894    

Page 16: PI3$Kinases:,An,Example,of,Holo$WaterMap, Modelling,content.schrodinger.com/Training+Material/WM/PI3K/PI3-Kinases_fi… · Perplexing,SAR:,Stereochemistry,Controls,AcKvity,and, SelecKvity

ConKnuous-­‐WaterMaps  and  Holo-­‐WaterMap    •  In  general,  finding  corresponding  hydraTon-­‐

sites  between  two  WaterMaps  is  not  easy  –  Frequently  the  changes  between  WaterMaps  are  

too  great  –  Manually  pairing  up  hydraTon-­‐sites  between  

WaterMaps  is  highly  tedious  and  error  prone  –  There  are  also  occasional  ‘edge  effects’  

•  A  subtle  movement  of  hydraTon-­‐sites  can  mean  that  the  N-­‐closest  sites  to  one  ligand  are  not  the  N-­‐closest  sites  to  another  

•  ConTnuous-­‐WaterMaps  have  none  of  these  problems  –  A  point  (x,y,z)  in  one  conTnuous-­‐WaterMap  is  

comparable  to  the  same  point  in  a  second  conTnuous-­‐WaterMap  •  All  that  is  required  is  a  good  alignment    

–  Defining  the  region  of  interest  is  much  easier  •  A  simple  ‘distance  from  the  ligand’  can  be  used,  rather  

than  having  to  make  judgement  calls  on  what  hydraTon-­‐sites  should  and  shouldn’t  be  included  

Page 17: PI3$Kinases:,An,Example,of,Holo$WaterMap, Modelling,content.schrodinger.com/Training+Material/WM/PI3K/PI3-Kinases_fi… · Perplexing,SAR:,Stereochemistry,Controls,AcKvity,and, SelecKvity

Aside  –  The  GeneraKon  of  ConKnuous  WaterMaps  •  The  generaTon  of  conTnuous  WaterMaps  follows  basically  the  same  procedure,  however  the  clustering  stage  is  removed  and  replaced  with  a  set  of  calculaTons  on  a  high-­‐resoluTon  grid  (0.5Å)  covering  the  region  of  interest:  – The  thermodynamic  properTes  are  calculated  for  each  lamce  point  

•  The  enthalpies  are  averaged  over  water-­‐molecules  found  inside  each  cell  •  The  entropies  are  calculated  for  all  water  molecules  within  a  1Å  sphere,  centered  on  each  lamce  point  –  30o  bins  are  used  for  rotaTonal  entropy  calculaTon  

 

Page 18: PI3$Kinases:,An,Example,of,Holo$WaterMap, Modelling,content.schrodinger.com/Training+Material/WM/PI3K/PI3-Kinases_fi… · Perplexing,SAR:,Stereochemistry,Controls,AcKvity,and, SelecKvity

CreaKng  a  Custom  Analysis  Script  for  Holo-­‐WaterMap  Analysis  

•  The  conTnuous-­‐WaterMap  data  is  saved  as  a  3D-­‐array  in  a  .CNS  format  file  (stored  within  the  output  .zip  archive  of  a  WaterMap  calculaTon)  –  The  Schrödinger  Python  API  has  a  specific  

module  dedicated  to  loading/creaTng,  manipulaTng  and  saving  such  data  •  This  module  is  known  as  schrodinger.analysis.  visanalysis  

•  In  addiTon  to  being  useful  for  analysing  conTnuous-­‐WaterMap  data,  the  module  can  also  be  used  to  process  data  such  as  Jaguar  ESP/electron-­‐density  informaTon  and  SiteMap  grids  –  The  API’s  documentaTon  gives  several  

examples  of  its  use  

schrodinger.analysis.visanalysis  

volumedata    

Implements  the  VolumeData  class,  which  represents  3D-­‐volume  information  e.g.  SiteMap  grids,  electrostatic  potentials,  ‘densities’  

volumedataio    

Allows  VolumeData  objects  to  be  created  from  various  file  formats.  Permits  VolumeData  objects  

to  be  saved  to  new  .vis  and  .ccp4  files.    

volumedatautils    Various  utility  functions.  Includes,  bounding-­‐box  calculations,  interpolation  and  fast  grid-­‐

point  location  facilities.  

Page 19: PI3$Kinases:,An,Example,of,Holo$WaterMap, Modelling,content.schrodinger.com/Training+Material/WM/PI3K/PI3-Kinases_fi… · Perplexing,SAR:,Stereochemistry,Controls,AcKvity,and, SelecKvity

Aside:  The  Underlying  Code  Load  .CNS  file  containing  the  conTnuous  holo-­‐WaterMap.  This  returns  a  VolumeData  instance:    holoWM  =  volumedataio.LoadCNSFile(filenameWM)  

Load  the  molecule  that  defines  the  integraTon  volume.  molecule  is  an  instance  of  the    Schrödinger  structure  class:    reader  =  structure.StructureReader(filenameStr)  molecule  =  reader.next()  

Locate  the  points  in  the  holo-­‐WaterMap  that  are  within  X(Å)  of  the  nearest  ligand  atom.  The  DataPointLocator  class  allows  for  fast  distance-­‐based  lookup:    dpl  =  volumedatautils.DataPointLocator(holoWM)  for  at  in  molecule.atom:          dpl.SearchForDataPointsWithin(  

   [at.x,  at.y,  at.z],  integrationRadius)  dpl.UniquifyResults()  

Perform  the  integraTon  (summaTon):    acc  =  0.0  for  x,  y,  z  in  dpl.Results:          acc  +=  holoWM.getData()[x][y][z]    print  “Accumulated  WM  =  %f”  %  acc    That’s  it!  

Page 20: PI3$Kinases:,An,Example,of,Holo$WaterMap, Modelling,content.schrodinger.com/Training+Material/WM/PI3K/PI3-Kinases_fi… · Perplexing,SAR:,Stereochemistry,Controls,AcKvity,and, SelecKvity

Aside:  Actually  Running  the  Custom  Code  •  Running  custom  code  is  very  easy:  – Firstly  we  ensure  that  the  custom  code  can  be  found  

•  Saving  it  in  the  working-­‐directory  is  a  good  idea  •  AlternaTvely,  add  it  to  the  PYTHONPATH  

– Extract  the  conTnuous-­‐WaterMap  .CNS  file  from  the  main  WaterMap  .zip  archive  

–  Invoke  the  custom  code  using  the  standard  $SCHRODINGER/run  command:    $SCHRODINGER/run  INTEGRATE_WM.py  <continuousWM.cns>  <ligand.mae>  <integration  radius>  

– The  results  are  simply  printed  on  stdout  

Page 21: PI3$Kinases:,An,Example,of,Holo$WaterMap, Modelling,content.schrodinger.com/Training+Material/WM/PI3K/PI3-Kinases_fi… · Perplexing,SAR:,Stereochemistry,Controls,AcKvity,and, SelecKvity

Results  

PI3KβR² = 0.7447

PI3KδR² = 0.8047

-13

-12

-11

-10

-9

-8

-7 12 14 16 18 20 22 24 26 28 30 32

ΔG B

ind  (kcal/m

ol)  

∫EnergyH2O  (arbitrary  units)  Results  

 The  results  from  the  conTnuous-­‐WaterMaps  show  a  very  nice  

(semi-­‐)quanTtaTve  relaTonship  with  the  observed  acTviTes.    

   That  we  can  explain  75%  and  80%  of  the  variance  of  binding-­‐energy  on  PI3K-­‐β  and  δ  respecTvely  is  quite  

remarkable  given  that  this  approach  uderly  neglects  any  effects  of  

protein-­‐ligand  interacTons  and/or  ligand  strain  terms.  It  does  however  

underline  the  importance  of  solvaTon-­‐effects  on  fully  understanding  SAR.    

Page 22: PI3$Kinases:,An,Example,of,Holo$WaterMap, Modelling,content.schrodinger.com/Training+Material/WM/PI3K/PI3-Kinases_fi… · Perplexing,SAR:,Stereochemistry,Controls,AcKvity,and, SelecKvity

A  Possible  Structural  Basis  

Page 23: PI3$Kinases:,An,Example,of,Holo$WaterMap, Modelling,content.schrodinger.com/Training+Material/WM/PI3K/PI3-Kinases_fi… · Perplexing,SAR:,Stereochemistry,Controls,AcKvity,and, SelecKvity

PI3K-­‐β  and  PI3K-­‐δ  Water-­‐Networks  

Holo-­‐WaterMap  PI3K-­‐β B(U)    

We  can  readily  trace  a  water-­‐network  from  856Asp  to  the  main  high-­‐energy  water-­‐molecule.  The  

orientaTons  of  the  waters  and  their  separaTons  make  for  a  convincing  network.  

Holo-­‐WaterMap  PI3K-­‐δ B(U)    

A  similar  network  exists  in  PI3K-­‐δ  between  the  high-­‐energy  water  and  836Asn.  The  network  is  visually  convincing,  but  

doesn’t  appear  as  Tghtly  bound.  Something  that  is  borne  out  by  the  enthalpy/entropy  values  calculated  by  WaterMap.  

Page 24: PI3$Kinases:,An,Example,of,Holo$WaterMap, Modelling,content.schrodinger.com/Training+Material/WM/PI3K/PI3-Kinases_fi… · Perplexing,SAR:,Stereochemistry,Controls,AcKvity,and, SelecKvity

Examining  the  Water-­‐Network  with  Pure  Molecular-­‐Dynamics  

•  The  water-­‐networks  derived  from  the  WaterMap  results  are  visually  quite  convincing  

•  We  can  further  invesTgate  these  water-­‐networks  via  a  normal  MD-­‐simulaTon*  –  An  unrestrained  simulaTon  removes  the  necessary,  but  arTficial,  restraints  created  by  WaterMap  and  allows  us  to  esTmate  the  water-­‐structure  in  the  presence  of  general  thermal-­‐moTons  

•  The  water-­‐density  can  visualised  by  a  simple  script  –  The  reliability  of  this  ‘mean  water-­‐density’  is  enTrely  dependent  on  the  alignment  of  the  various  snapshots  •  Bad  alignments  and/or  excessive  flexibility  makes  a  simple  spaTal-­‐averaging  somewhat  ill-­‐defined  

–  In  this  case  things  aren’t  too  bad  

*20ns,  1atm,  300K,  NPT  simulaTon  

Page 25: PI3$Kinases:,An,Example,of,Holo$WaterMap, Modelling,content.schrodinger.com/Training+Material/WM/PI3K/PI3-Kinases_fi… · Perplexing,SAR:,Stereochemistry,Controls,AcKvity,and, SelecKvity

MD-­‐Derived  Water-­‐Density  

MD-­‐Derived  Water-­‐Density  PI3K-­‐β B(U)    

The  MD-­‐derived  water-­‐network  is  in  good  agreement  with  the  PI3K-­‐β  results  from  WaterMap.  A  slight  difference  in  the  

MD-­‐results  is  that  852Glu  has  moved  to  assist  856Asp  in  forming  the  network  to  the  trapped  water-­‐molecule.  

MD-­‐Derived  Water-­‐Density  PI3K-­‐δ B(U)    

The  MD  and  WaterMap  results  also  broadly  agree  in  the  case  of  PI3K-­‐δ.  However,  MD  suggests  a  diminished  role  for  836Asn  in  supporTng  the  water-­‐network.  Leaving    832Asp  to  act  as  its  anchor.  The  network  is  clearly  more  sparse  than  in  PI3K-­‐β.  

60%  

70%  

80%  

%Occ  

Page 26: PI3$Kinases:,An,Example,of,Holo$WaterMap, Modelling,content.schrodinger.com/Training+Material/WM/PI3K/PI3-Kinases_fi… · Perplexing,SAR:,Stereochemistry,Controls,AcKvity,and, SelecKvity

WRAPPING  UP  

Page 27: PI3$Kinases:,An,Example,of,Holo$WaterMap, Modelling,content.schrodinger.com/Training+Material/WM/PI3K/PI3-Kinases_fi… · Perplexing,SAR:,Stereochemistry,Controls,AcKvity,and, SelecKvity

Acknowledgments  Many  thanks  to:    The  team  at  Sanofi:    Thomas  Bertrand,  Jean-­‐Christophe  Carry,  Frank  Halley,  Andreas  Karlsson,  Magali  Mathieu,  Hervé  Minoux,  Marc-­‐Antoine  Perrin  Benoit  Robert  and  Laurent  Schio  

   Schrodinger:    Woody  Sherman  

Page 28: PI3$Kinases:,An,Example,of,Holo$WaterMap, Modelling,content.schrodinger.com/Training+Material/WM/PI3K/PI3-Kinases_fi… · Perplexing,SAR:,Stereochemistry,Controls,AcKvity,and, SelecKvity

PI3-­‐Kinases:  An  Example  of  Holo-­‐WaterMap  Modelling  

Daniel  D.  Robinson  [email protected]