Biomedical Engineering Reference
In-Depth Information
These attractors become the 8 gene expressions used as input to our method and
All-SAT on the CNF results in 72 possible satisfying BNs as the output. Furthermore,
we observe that one of the 72 BNs has the correct logic function for the
p53
network. If
we count the number of selected functions per gene, we find that for
ATM
,
p
53,
Wip
,
and
Mdm
2, there are 6, 4, 1, and 3 functions respectively. These results can help
biologists tune their experiments to understand gene regulatory function. Also, in-
formation from other studies (such as curated data or pathway information) can be
used to further reduce the number of solution BNs.
3.5
Chapter Summary
In this chapter, we have presented an efficient and general SAT-based method for
determining logic functions from gene expression data. Our approach implicitly
explores all possible logic functions for each gene based on the predictor set, and se-
lects functions that match the gene expression observations using a SAT formulation.
Each SAT solution is a Boolean network, and the results of our method is a family
of BNs that match the predictor set and gene expressions. Our SAT-based method
is validated on two GRNs and demonstrates the importance of gene expression data
to constraining the number of satisfying BNs. We also test the method on the
p53
network and show how our results can be used to find the gene functions. Due to its
generality and efficiency, this algorithm can easily be extended to large networks,
and can be augmented to utilize gene expression data from multiple sources (curated
pathway information, for example).
Thus far, our inference of the GRN predictor set and regulating function has
been done using binary valued gene expression data. In practice, gene expressions
are initially measured as continuous values, from which the values are converted to
binary values. While binary gene expressions simplify our analysis, continuous gene
expression may provide a richer and more detailed observation of the gene state and
GRN. The next chapter explores methods to infer the GRN from continuous gene
expression data.
References
1. Guelzim, N. et al. : Topological and causal structure of the yeast transcriptional regulatory
network. Nat. Genet.,
31
, 60-63, (2002)
2. Kauffman, S. A.: Metabolic stability and epigenesis in randomly constructed genetic nets.
J.Theor.Biol..
22
(3), 437-467, (1969)
3. Lin, P. K., Khatri S.P., Determining gene function in Boolean networks using Boolean sat-
isfiability, in Genomic Signal Processing and Statistics (GENSIPS), 2012 IEEE International
Workshop on. IEEE, pp. 1-4. (2012)
4. Layek, R., Datta, A., Dougherty, E.R.: From biological pathways to regulatory networks,
Decision and Control (CDC), 2010 49th IEEE Conference on, pp. 5781-5786. (2010)
5. Layek, R., Datta, A., Bittner, M., Dougherty, E.R.: Cancer therapy design based on pathway
logic. Bioinform.
27
(4), 548-555 (2011)