Predicting Long Term behavior of Genetic Regulatory Networks with Answer Set Programming

[email protected] PREDICTING LONG TERM BEHAVIOR OF GENETIC REGULATORY NETWORKS WITH
ASP Gustavo Torres

of 27 GENETIC REGULATORY NETWORKS (GRNs) 2 ASP for Long
Term Behavior of GRNs GENE 1 GENE 3 GENE 2 mRNA 1 mRNA 2 mRNA 3 Protein 1 Protein 2 Protein 3 GRN

of 27 GENETIC REGULATORY NETWORKS (GRNs) 2 ASP for Long
Term Behavior of GRNs GENE 1 GENE 3 GENE 2 mRNA 1 mRNA 2 mRNA 3 Protein 1 Protein 2 Protein 3 GRN g1 g3 g3 inhibits g1 g2 activates g3 g2 BOOLEAN NETWORK g1, g2, g3 ∈ {0,1}

of 27 WHY BOOLEAN NETWORKS? 3 ASP for Long Term
Behavior of GRNs Simple enough to be scalable Detailed biological information is not often available Complex behavior can come from simple structure

of 27 ASP for Long Term Behavior of GRNs ANSWER
SET PROGRAMMING 4 Boolean Networks are deﬁned according to a set of rules that control its dynamics E.g. if an inhibitor is present the gene will not be switched on in the next time step We rely on ASP, a rule based declarative framework, to ﬂexibly describe the model

SET PROGRAMMING 5 g2 g1 g3 .n g1 3 g1 g2 g3 - - 1 0 1 - 0 1 - 1 0 1 0 0 0 0 SAT ASP activates(g1,g1) inhibits(g3,g1) activates(g2,g1)

SET PROGRAMMING 5 g2 g1 g3 .n g1 3 g1 g2 g3 - - 1 0 1 - 0 1 - 1 0 1 0 0 0 0 SAT ASP activates(g1,g1) inhibits(g3,g1) activates(g2,g1) inhibited(X,T+1) inhibits(Y,X) active(Y,T)

SET PROGRAMMING 6 g2 g1 g3 .n g1 3 g1 g2 g3 - - 1 0 1 - 0 1 - 1 0 1 0 0 0 0 SAT ASP activates(g1,g1) inhibits(g3,g1) activates(g2,g1) inhibited(X,T+1) inhibits(Y,X) active(Y,T)

of 27 ASP for Long Term Behavior of GRNs APPLICABILITY
OF ASP 7 To show the use of ASP for Boolean Networks modeling we pursued the problem of calculating the attractors of the modeled dynamic system An attractor is a set of states in which the dynamics of the network cycles indefinitely These attractor states usually correspond to important biological decisions

of 27 DYNAMICS OF GRNs: SYNCHRONOUS 8 g2 g1 g3
100 111 110 001 010 101 011 000 g1=0 g2=1 g3=1 Boolean Network State Transition Graph (STG) ASP for Long Term Behavior of GRNs

100 111 110 001 010 101 011 000 ASP for Long Term Behavior of GRNs

of 27 DYNAMICS OF GRNs: ATTRACTORS 14 g2 g1 g3

of 27 CALCULATING ALL ATTRACTORS 15 g2 g1 g3 100
111 110 001 010 101 011 000 ASP for Long Term Behavior of GRNs Pick a path in the STG of size 2

111 110 001 010 101 011 000 ASP for Long Term Behavior of GRNs Pick a path in the STG of size 4

111 110 001 010 101 011 000 ASP for Long Term Behavior of GRNs Find cycles inside the path

001 011 000 ASP for Long Term Behavior of GRNs Pick a path in the STG of size 4

of 27 EXPERIMENTAL RESULTS 19 ASP for Long Term Behavior
of GRNs T helper (Th) cells are a subgroup of lymphocytes important for the immune system Divided into Th1 and Th2 cells each responsible for different diseases Pedicini et al (2010) used gene knockout experiments to tell whether Th1 and Th2 are mutually exclusive

of 27 EXPERIMENTAL RESULTS 20 ASP for Long Term Behavior
of GRNs 0 5,00 10,00 15,00 20,00 COT GATA3 IKBKB IRAK IRF4 JAK1 JAK3 LCK MAF NFAT NFKB Solving time for the Th Cell Helper network (Pedicini et al (2010)) SAT ASP

of 27 ASYNCHRONOUS BEHAVIOR 21 ASP for Long Term Behavior
of GRNs In addition to SAT, which only implements ﬁnding attractors for the synchronous updated network we also considered asynchronously updated networks. An asynchronous updated network considers the fact that genes do not change their state all at the same moment

of 27 EXTENDING ATTRACTORS TO ASYNCHRONOUS 22 g2 g1 g3

of 27 DYNAMICS OF GRNs: ASYNCHRONOUS 23 g2 g1 g3

of 27 CALCULATING ALL ATTRACTORS: ALGORITHM 24 changed ( X,
T ) active ( X, T ) , inhibited ( X, T 1) , protein ( X ) , T > 0 . changed ( X, T ) inhibited ( X, T ) , active ( X, T 1) , protein ( X ) , T > 0 . N = # changed ( X, T ) , N 2 , protein ( X ) , T > 0 . active(g1, T), inhibited(g2, T), active(g3, T). Asynchronous 101 Do not repeat visited states ASP for Long Term Behavior of GRNs 001 000 011 010 111 110 101 100

of 27 CALCULATING ALL ATTRACTORS: ALGORITHM 25 changed ( X,
T ) active ( X, T ) , inhibited ( X, T 1) , protein ( X ) , T > 0 . changed ( X, T ) inhibited ( X, T ) , active ( X, T 1) , protein ( X ) , T > 0 . N = # changed ( X, T ) , N 2 , protein ( X ) , T > 0 . active(g1, T), inhibited(g2, T), active(g3, T). Asynchronous 101 Do not repeat visited states ASP for Long Term Behavior of GRNs 000 011 110

of 27 CONCLUSIONS ‣ ASP provides a natural way to
encode GRNs’ behavior and adds flexibility ‣ Attractors can be used as predictors of long term behavior of GRNs ‣ The running time for calculating all attractors becomes proportional to the number of attractors rather than in the size of the network (2n) 26 ASP for Long Term Behavior of GRNs

of 27 ACKNOWLEDGMENTS ‣ Kathleen Marchal, Yves Van de Peer,
Martine De Cock ‣ Dubrova, Teslenko (2011) A SAT based algorithm for finding Attractors in Synchronous Boolean Networks ‣ Pedicini, Barrenäs, Clancy, Castiglione, Hovig, Kanduri, Santoni et al (2010) Combining network modeling and gene expression microarray analysis to explore the dynamics of Th1 and Th2 cell regulation 27 ASP for Long Term Behavior of GRNs

Predicting Long Term behavior of Genetic Regulatory Networks with Answer Set Programming

Predicting Long Term behavior of Genetic Regulatory Networks with Answer Set Programming

Gustavo Torres

More Decks by Gustavo Torres

Other Decks in Research

Featured

Transcript

[email protected] PREDICTING LONG TERM BEHAVIOR OF GENETIC REGULATORY NETWORKS WITH

of 27 GENETIC REGULATORY NETWORKS (GRNs) 2 ASP for Long

of 27 GENETIC REGULATORY NETWORKS (GRNs) 2 ASP for Long

of 27 WHY BOOLEAN NETWORKS? 3 ASP for Long Term

of 27 ASP for Long Term Behavior of GRNs ANSWER

of 27 ASP for Long Term Behavior of GRNs ANSWER

of 27 ASP for Long Term Behavior of GRNs ANSWER

of 27 ASP for Long Term Behavior of GRNs ANSWER

of 27 ASP for Long Term Behavior of GRNs APPLICABILITY

of 27 DYNAMICS OF GRNs: SYNCHRONOUS 8 g2 g1 g3

of 27 DYNAMICS OF GRNs: SYNCHRONOUS 9 g2 g1 g3

of 27 DYNAMICS OF GRNs: SYNCHRONOUS 10 g2 g1 g3

of 27 DYNAMICS OF GRNs: SYNCHRONOUS 11 g2 g1 g3

of 27 DYNAMICS OF GRNs: SYNCHRONOUS 12 g2 g1 g3

of 27 DYNAMICS OF GRNs: SYNCHRONOUS 13 g2 g1 g3

of 27 DYNAMICS OF GRNs: ATTRACTORS 14 g2 g1 g3

of 27 CALCULATING ALL ATTRACTORS 15 g2 g1 g3 100

of 27 CALCULATING ALL ATTRACTORS 16 g2 g1 g3 100

of 27 CALCULATING ALL ATTRACTORS 17 g2 g1 g3 100

of 27 CALCULATING ALL ATTRACTORS 18 g2 g1 g3 100

of 27 EXPERIMENTAL RESULTS 19 ASP for Long Term Behavior

of 27 EXPERIMENTAL RESULTS 20 ASP for Long Term Behavior

of 27 ASYNCHRONOUS BEHAVIOR 21 ASP for Long Term Behavior

of 27 EXTENDING ATTRACTORS TO ASYNCHRONOUS 22 g2 g1 g3

of 27 DYNAMICS OF GRNs: ASYNCHRONOUS 23 g2 g1 g3

of 27 CALCULATING ALL ATTRACTORS: ALGORITHM 24 changed ( X,

of 27 CALCULATING ALL ATTRACTORS: ALGORITHM 25 changed ( X,

of 27 CONCLUSIONS ‣ ASP provides a natural way to

of 27 ACKNOWLEDGMENTS ‣ Kathleen Marchal, Yves Van de Peer,