Limits...
MAE-FMD: multi-agent evolutionary method for functional module detection in protein-protein interaction networks.

Ji JZ, Jiao L, Yang CC, Lv JW, Zhang AD - BMC Bioinformatics (2014)

Bottom Line: Next, it applies several evolutionary operators, i.e., competition, crossover, and mutation, to realize information exchange among agents as well as solution evolution.Experimental results show that the approach is more effective compared to several other existing algorithms.The algorithm has the characteristics of outstanding recall, F-measure, sensitivity and accuracy while keeping other competitive performances, so it can be applied to the biological study which requires high accuracy.

View Article: PubMed Central - PubMed

Affiliation: College of Computer Science, Beijing University of Technology, Chaoyang District, Beijing, China. jjz01@bjut.edu.cn.

ABSTRACT

Background: Studies of functional modules in a Protein-Protein Interaction (PPI) network contribute greatly to the understanding of biological mechanisms. With the development of computing science, computational approaches have played an important role in detecting functional modules.

Results: We present a new approach using multi-agent evolution for detection of functional modules in PPI networks. The proposed approach consists of two stages: the solution construction for agents in a population and the evolutionary process of computational agents in a lattice environment, where each agent corresponds to a candidate solution to the detection problem of functional modules in a PPI network. First, the approach utilizes a connection-based encoding scheme to model an agent, and employs a random-walk behavior merged topological characteristics with functional information to construct a solution. Next, it applies several evolutionary operators, i.e., competition, crossover, and mutation, to realize information exchange among agents as well as solution evolution. Systematic experiments have been conducted on three benchmark testing sets of yeast networks. Experimental results show that the approach is more effective compared to several other existing algorithms.

Conclusions: The algorithm has the characteristics of outstanding recall, F-measure, sensitivity and accuracy while keeping other competitive performances, so it can be applied to the biological study which requires high accuracy.

Show MeSH
The effects of filter thresholdδon 6 performance metrics.(a) reveals the relation between the δ value and recall, F-measure and precision; and (b) displays the relation between the δ value and sensitivity, accuracy and PPV.
© Copyright Policy - open-access
Related In: Results  -  Collection

License 1 - License 2
getmorefigures.php?uid=PMC4262229&req=5

Fig10: The effects of filter thresholdδon 6 performance metrics.(a) reveals the relation between the δ value and recall, F-measure and precision; and (b) displays the relation between the δ value and sensitivity, accuracy and PPV.

Mentions: Figure 10 gives the effects of filter threshold δ on 6 performance metrics. As shown in Figure 10(a), the recall and F-measure have a similar trend, namely, their values slowly increase as δ increases at the beginning, then gently decrease after δ gets to 0.12. However, the rate of change is slightly different for the two metrics where the values of recall have larger changes than those of F-measure. Meanwhile, the precision maintains a relatively stable value around 0.45 though there are two small peaks at δ=0.04 and 0.12. Figure 10(b) investigates the relationship between δ and PPV, the accuracy and the sensitivity. As δ increases, three metrics have different tendencies. In detail, the sensitivity obviously decreases from 0.75 to 0.52, the PPV increases from 0.30 to 0.32 when δ locates in [0.02, 0.16], then keeps a larger value (0.32) when δ>0.16 while the accuracy holds steady at 0.46 when δ varies from 0.02 to 0.14, then slightly decreases from 0.46 to 0.41 when δ locates in [0.14, 0.2]. The main reason for these different trends is that only those modules whose similarity is strong enough are merged along with the value of δ increasing, thus making the number of clusters to increase and the average size of a cluster to be small. To make a balance, we employ δ=0.12 in our algorithm.Figure 10


MAE-FMD: multi-agent evolutionary method for functional module detection in protein-protein interaction networks.

Ji JZ, Jiao L, Yang CC, Lv JW, Zhang AD - BMC Bioinformatics (2014)

The effects of filter thresholdδon 6 performance metrics.(a) reveals the relation between the δ value and recall, F-measure and precision; and (b) displays the relation between the δ value and sensitivity, accuracy and PPV.
© Copyright Policy - open-access
Related In: Results  -  Collection

License 1 - License 2
Show All Figures
getmorefigures.php?uid=PMC4262229&req=5

Fig10: The effects of filter thresholdδon 6 performance metrics.(a) reveals the relation between the δ value and recall, F-measure and precision; and (b) displays the relation between the δ value and sensitivity, accuracy and PPV.
Mentions: Figure 10 gives the effects of filter threshold δ on 6 performance metrics. As shown in Figure 10(a), the recall and F-measure have a similar trend, namely, their values slowly increase as δ increases at the beginning, then gently decrease after δ gets to 0.12. However, the rate of change is slightly different for the two metrics where the values of recall have larger changes than those of F-measure. Meanwhile, the precision maintains a relatively stable value around 0.45 though there are two small peaks at δ=0.04 and 0.12. Figure 10(b) investigates the relationship between δ and PPV, the accuracy and the sensitivity. As δ increases, three metrics have different tendencies. In detail, the sensitivity obviously decreases from 0.75 to 0.52, the PPV increases from 0.30 to 0.32 when δ locates in [0.02, 0.16], then keeps a larger value (0.32) when δ>0.16 while the accuracy holds steady at 0.46 when δ varies from 0.02 to 0.14, then slightly decreases from 0.46 to 0.41 when δ locates in [0.14, 0.2]. The main reason for these different trends is that only those modules whose similarity is strong enough are merged along with the value of δ increasing, thus making the number of clusters to increase and the average size of a cluster to be small. To make a balance, we employ δ=0.12 in our algorithm.Figure 10

Bottom Line: Next, it applies several evolutionary operators, i.e., competition, crossover, and mutation, to realize information exchange among agents as well as solution evolution.Experimental results show that the approach is more effective compared to several other existing algorithms.The algorithm has the characteristics of outstanding recall, F-measure, sensitivity and accuracy while keeping other competitive performances, so it can be applied to the biological study which requires high accuracy.

View Article: PubMed Central - PubMed

Affiliation: College of Computer Science, Beijing University of Technology, Chaoyang District, Beijing, China. jjz01@bjut.edu.cn.

ABSTRACT

Background: Studies of functional modules in a Protein-Protein Interaction (PPI) network contribute greatly to the understanding of biological mechanisms. With the development of computing science, computational approaches have played an important role in detecting functional modules.

Results: We present a new approach using multi-agent evolution for detection of functional modules in PPI networks. The proposed approach consists of two stages: the solution construction for agents in a population and the evolutionary process of computational agents in a lattice environment, where each agent corresponds to a candidate solution to the detection problem of functional modules in a PPI network. First, the approach utilizes a connection-based encoding scheme to model an agent, and employs a random-walk behavior merged topological characteristics with functional information to construct a solution. Next, it applies several evolutionary operators, i.e., competition, crossover, and mutation, to realize information exchange among agents as well as solution evolution. Systematic experiments have been conducted on three benchmark testing sets of yeast networks. Experimental results show that the approach is more effective compared to several other existing algorithms.

Conclusions: The algorithm has the characteristics of outstanding recall, F-measure, sensitivity and accuracy while keeping other competitive performances, so it can be applied to the biological study which requires high accuracy.

Show MeSH