Format

Send to

Choose Destination
BMC Syst Biol. 2017 Sep 21;11(Suppl 4):81. doi: 10.1186/s12918-017-0454-9.

Bayesian network model for identification of pathways by integrating protein interaction with genetic interaction data.

Author information

1
School of Mathematics and Computational Science, Xiangtan University, Xiangtan, 411105, China. fuch@synu.edu.cn.
2
School of Mathematics and System Science, Shenyang Normal University, Shenyang, 110034, China. fuch@synu.edu.cn.
3
School of Mathematics and System Science, Shenyang Normal University, Shenyang, 110034, China.
4
Center of Systems Biology and Bioinformatics, Wake Forest School of Medicine, Winston-Salem, NC, 27157, USA.
5
School of Mathematics and Computational Science, Xiangtan University, Xiangtan, 411105, China. yuzuguo@aliyun.com.

Abstract

BACKGROUND:

Molecular interaction data at proteomic and genetic levels provide physical and functional insights into a molecular biosystem and are helpful for the construction of pathway structures complementarily. Despite advances in inferring biological pathways using genetic interaction data, there still exists weakness in developed models, such as, activity pathway networks (APN), when integrating the data from proteomic and genetic levels. It is necessary to develop new methods to infer pathway structure by both of interaction data.

RESULTS:

We utilized probabilistic graphical model to develop a new method that integrates genetic interaction and protein interaction data and infers exquisitely detailed pathway structure. We modeled the pathway network as Bayesian network and applied this model to infer pathways for the coherent subsets of the global genetic interaction profiles, and the available data set of endoplasmic reticulum genes. The protein interaction data were derived from the BioGRID database. Our method can accurately reconstruct known cellular pathway structures, including SWR complex, ER-Associated Degradation (ERAD) pathway, N-Glycan biosynthesis pathway, Elongator complex, Retromer complex, and Urmylation pathway. By comparing N-Glycan biosynthesis pathway and Urmylation pathway identified from our approach with that from APN, we found that our method is able to overcome its weakness (certain edges are inexplicable). According to underlying protein interaction network, we defined a simple scoring function that only adopts genetic interaction information to avoid the balance difficulty in the APN. Using the effective stochastic simulation algorithm, the performance of our proposed method is significantly high.

CONCLUSION:

We developed a new method based on Bayesian network to infer detailed pathway structures from interaction data at proteomic and genetic levels. The results indicate that the developed method performs better in predicting signaling pathways than previously described models.

KEYWORDS:

Bayesian model; Biological pathway; Genetic interaction; Protein interaction

PMID:
28950903
PMCID:
PMC5615243
DOI:
10.1186/s12918-017-0454-9
[Indexed for MEDLINE]
Free PMC Article

Supplemental Content

Full text links

Icon for BioMed Central Icon for PubMed Central
Loading ...
Support Center