integrating mathematics and machine learning for drug design...drug design & discovery resource...

14
Guowei Wei Mathematics Biochemistry & Molecular biology Michigan State University http://www.math.msu.edu/~wei D3R 2018 Workshop February 22-23, La Jolla Grant support: NSF, NIH, MSU and BMS Integrating Mathematics and Machine Learning for Drug Design

Upload: others

Post on 25-May-2020

8 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Integrating Mathematics and Machine Learning for Drug Design...Drug Design & Discovery Resource (D3R) Grand Challenge 2 Given: Farnesoid X receptor (FXR) and 102 ligands Tasks: Dock

Guowei WeiMathematics

Biochemistry & Molecular biologyMichigan State University

http://www.math.msu.edu/~wei

D3R 2018 WorkshopFebruary 22-23, La Jolla

Grant support:NSF, NIH, MSU and BMS

Integrating Mathematics and Machine Learning for Drug Design

Page 2: Integrating Mathematics and Machine Learning for Drug Design...Drug Design & Discovery Resource (D3R) Grand Challenge 2 Given: Farnesoid X receptor (FXR) and 102 ligands Tasks: Dock
Page 3: Integrating Mathematics and Machine Learning for Drug Design...Drug Design & Discovery Resource (D3R) Grand Challenge 2 Given: Farnesoid X receptor (FXR) and 102 ligands Tasks: Dock

Möbius Strips (1858)

Klein Bottle (1882)

Classical topological objects

Torus

Double Torus

Sphere

Trefoil Knot

Seven Bridges of Königsberg

Leonhard Euler (1735)

Leonhard Paul Euler(Swiss Mathematician,

April 15, 1707 – Sept 18 1783)

Page 4: Integrating Mathematics and Machine Learning for Drug Design...Drug Design & Discovery Resource (D3R) Grand Challenge 2 Given: Farnesoid X receptor (FXR) and 102 ligands Tasks: Dock

Topological invariants: Betti numbersb0 is the number of connected components.b1 is the number of tunnels or circles.b2 is the number of cavities or voids.

Circle TorusPoint Sphere

001

2

1

0

===

bbb

011

2

1

0

===

bbb

101

2

1

0

===

bbb

121

2

1

0

===

bbb

Page 5: Integrating Mathematics and Machine Learning for Drug Design...Drug Design & Discovery Resource (D3R) Grand Challenge 2 Given: Farnesoid X receptor (FXR) and 102 ligands Tasks: Dock

Vietoris-Rips complexes of planar point setsSimplexes:

0-simplex 1-simplex 2-simplex 3-simplexSimplicial complexes of ten points:

Page 6: Integrating Mathematics and Machine Learning for Drug Design...Drug Design & Discovery Resource (D3R) Grand Challenge 2 Given: Farnesoid X receptor (FXR) and 102 ligands Tasks: Dock

( )kk

k

kk

kk

kk

HB

ZH

BZ

Rank

ImKer

1

=

=

¶=¶=

+

b

∂kσk = (−1)i

i=0

k

∑ v0,v1,...,vi,...,vk{ }

ki

iicså

Simplexes:

0-simplex 1-simplex 2-simplex 3-simplex

Boundary operator:

Frosini and Nandi (1999),Robins (1999),Edelsbrunner, Letscher and Zomorodian (2002), Edelsbrunner and Harer, (2007)Kaczynski, Mischaikow and Mrozek (2004),Zomorodian and Carlsson (2005),Ghrist (2008),……k-chain:

Chain group: )( 2K,ZCk

Topological modeling - Persistent homology

Page 7: Integrating Mathematics and Machine Learning for Drug Design...Drug Design & Discovery Resource (D3R) Grand Challenge 2 Given: Farnesoid X receptor (FXR) and 102 ligands Tasks: Dock

Filtration, Vietoris-Rips complexes, and persistent barcodes (Xia, Wei, 2014)

Page 8: Integrating Mathematics and Machine Learning for Drug Design...Drug Design & Discovery Resource (D3R) Grand Challenge 2 Given: Farnesoid X receptor (FXR) and 102 ligands Tasks: Dock

Topological fingerprints of an alpha helix

(Xia & Wei, IJNMBE,

2014, 2015)

Page 9: Integrating Mathematics and Machine Learning for Drug Design...Drug Design & Discovery Resource (D3R) Grand Challenge 2 Given: Farnesoid X receptor (FXR) and 102 ligands Tasks: Dock

Time

2D persistence in protein 1UBQ unfolding

(Xia & Wei, JCC, 2015)

log10(N)

Rad

ius

0b

2b

1b

Page 10: Integrating Mathematics and Machine Learning for Drug Design...Drug Design & Discovery Resource (D3R) Grand Challenge 2 Given: Farnesoid X receptor (FXR) and 102 ligands Tasks: Dock

Topological learning based predictionsPrediction correlations for 2648 mutations on globular proteins

Prediction correlations for 223 mutations on membrane proteins

Prediction RMSD of LogP(Star set)

Classification of ligands & decoysDUD database 128,374 protein-ligand/decoy pairs

Binding affinity prediction of PDBBindv2013 core set of 195 complexes

Cang and Wei, PLOS CB,2017

Cang and Wei, PLOS CB,2017

Cang and Wei, PLOS CB,2018

Wu and Wei, JCC,2018

Page 11: Integrating Mathematics and Machine Learning for Drug Design...Drug Design & Discovery Resource (D3R) Grand Challenge 2 Given: Farnesoid X receptor (FXR) and 102 ligands Tasks: Dock

Drug Design & Discovery Resource (D3R) Grand Challenge 2 Given:Farnesoid Xreceptor(FXR)and102ligandsTasks:Dock102ligandstoFXR,andcomputetheirposes,bindingfreeenergiesandenergyranking

Dr Duc Nguyen

Page 12: Integrating Mathematics and Machine Learning for Drug Design...Drug Design & Discovery Resource (D3R) Grand Challenge 2 Given: Farnesoid X receptor (FXR) and 102 ligands Tasks: Dock

D3R Grand Challenge 2 Given:Farnesoid Xreceptor(FXR)and102ligandsTasks:Dock102ligandstoFXR,andcomputetheirposes,bindingfreeenergiesandenergyranking

Dr Duc Nguyen

Page 13: Integrating Mathematics and Machine Learning for Drug Design...Drug Design & Discovery Resource (D3R) Grand Challenge 2 Given: Farnesoid X receptor (FXR) and 102 ligands Tasks: Dock

Original protein-ligand Complex

Classify atoms into element

specific groups

Generate topological fingerprints

Multichannel images

(54x200)

Convolutional deep learning neural

network

Topological convolutional deep Learning architecture

Convolution (128x200)

Pooling (128x100)

Flattening (1xN)

Prediction

(Cang & Wei, PLOS CB, 2017)

Page 14: Integrating Mathematics and Machine Learning for Drug Design...Drug Design & Discovery Resource (D3R) Grand Challenge 2 Given: Farnesoid X receptor (FXR) and 102 ligands Tasks: Dock

D3R Grand Challenge 3Preliminary Evaluations, Subject to Revision and RefinementCathepsin Stage 1 Pose Predictions (partials) Scoring (partials)Free Energy SetsCathepsin Stage 1B Pose PredictionCathepsin Stage 2 Scoring (partials)Free Energy Sets

VEGFR2 Scoring (partials) JAK SC2 Scoring (partials) p38-α Scoring (partials)

JAK SC3 Free Energy Sets

TIE2 Scoring (partials)Free Energy Set 1Free Energy Set 2

ABL1 Scoring

1st

Dr Duc NguyenZixuan Cang