Limits...
Detecting protein complexes in protein interaction networks using a ranking algorithm with a refined merging procedure.

Hanna EM, Zaki N - BMC Bioinformatics (2014)

Bottom Line: It is important since it allows better understanding of cellular functions as well as malfunctions and it consequently leads to producing more effective cures for diseases.ProRank + was compared to several state-of-the-art approaches in order to show its effectiveness.The experimental results achieved by ProRank + show its ability to detect protein complexes in protein interaction networks.

View Article: PubMed Central - HTML - PubMed

Affiliation: College of Information Technology, United Arab Emirates University (UAEU, Al Ain 17551, United Arab Emirates. eileen.hanna@uaeu.ac.ae.

ABSTRACT

Background: Developing suitable methods for the identification of protein complexes remains an active research area. It is important since it allows better understanding of cellular functions as well as malfunctions and it consequently leads to producing more effective cures for diseases. In this context, various computational approaches were introduced to complement high-throughput experimental methods which typically involve large datasets, are expensive in terms of time and cost, and are usually subject to spurious interactions.

Results: In this paper, we propose ProRank+, a method which detects protein complexes in protein interaction networks. The presented approach is mainly based on a ranking algorithm which sorts proteins according to their importance in the interaction network, and a merging procedure which refines the detected complexes in terms of their protein members. ProRank + was compared to several state-of-the-art approaches in order to show its effectiveness. It was able to detect more protein complexes with higher quality scores.

Conclusions: The experimental results achieved by ProRank + show its ability to detect protein complexes in protein interaction networks. Eventually, the method could potentially identify previously-undiscovered protein complexes.The datasets and source codes are freely available for academic purposes at http://faculty.uaeu.ac.ae/nzaki/Research.htm.

Show MeSH
The detected complexes by the ProRank algorithm with the complex-overlap assumption when applied on the sub-network in Figure 2.
© Copyright Policy - open-access
Related In: Results  -  Collection

License 1 - License 2
getmorefigures.php?uid=PMC4230023&req=5

Figure 4: The detected complexes by the ProRank algorithm with the complex-overlap assumption when applied on the sub-network in Figure 2.

Mentions: Added to that, proteins could contribute in multiple cellular functions by being part of several protein complexes [13]. For instance, among the 1189 proteins contained in the MIPS catalog of protein complexes [22], 820 proteins (approx. 69%) belong to more than one complex. Similarly, among the 1279 covered by the SGD complex set [23], 332 proteins (approx. 26%) belong to multiple complexes. A protein interaction network is hence expected to comprise overlapping complexes, and accounting for this biological fact would most likely lead to more accurate complex-detection results. Accordingly, let us observe the effect of this adjustment on the protein-protein interaction network presented in Figure 2. The detected complexes, corresponding to applying the ProRank method with the added overlap assumption, are listed Figure 4. The results uphold the improvement added by the overlap extension which could potentially lead to a more correct detection of protein complexes. Actually, by allowing proteins to belong to more than one complex, the number of complexes formed from the identified essential proteins becomes higher indeed. However, it can be noticed that the amount of overlaps among some the detected complexes is relatively high. This was anticipated. Actually, since all essential proteins are now seeds for forming protein complexes, the ones that share numerous neighbors will certainly produce close and highly-overlapping protein complexes. In order to overcome this limitation and to further improve the quality of the predicted complexes, the following filtering and merging steps are added to the algorithm (Figure 5):


Detecting protein complexes in protein interaction networks using a ranking algorithm with a refined merging procedure.

Hanna EM, Zaki N - BMC Bioinformatics (2014)

The detected complexes by the ProRank algorithm with the complex-overlap assumption when applied on the sub-network in Figure 2.
© Copyright Policy - open-access
Related In: Results  -  Collection

License 1 - License 2
Show All Figures
getmorefigures.php?uid=PMC4230023&req=5

Figure 4: The detected complexes by the ProRank algorithm with the complex-overlap assumption when applied on the sub-network in Figure 2.
Mentions: Added to that, proteins could contribute in multiple cellular functions by being part of several protein complexes [13]. For instance, among the 1189 proteins contained in the MIPS catalog of protein complexes [22], 820 proteins (approx. 69%) belong to more than one complex. Similarly, among the 1279 covered by the SGD complex set [23], 332 proteins (approx. 26%) belong to multiple complexes. A protein interaction network is hence expected to comprise overlapping complexes, and accounting for this biological fact would most likely lead to more accurate complex-detection results. Accordingly, let us observe the effect of this adjustment on the protein-protein interaction network presented in Figure 2. The detected complexes, corresponding to applying the ProRank method with the added overlap assumption, are listed Figure 4. The results uphold the improvement added by the overlap extension which could potentially lead to a more correct detection of protein complexes. Actually, by allowing proteins to belong to more than one complex, the number of complexes formed from the identified essential proteins becomes higher indeed. However, it can be noticed that the amount of overlaps among some the detected complexes is relatively high. This was anticipated. Actually, since all essential proteins are now seeds for forming protein complexes, the ones that share numerous neighbors will certainly produce close and highly-overlapping protein complexes. In order to overcome this limitation and to further improve the quality of the predicted complexes, the following filtering and merging steps are added to the algorithm (Figure 5):

Bottom Line: It is important since it allows better understanding of cellular functions as well as malfunctions and it consequently leads to producing more effective cures for diseases.ProRank + was compared to several state-of-the-art approaches in order to show its effectiveness.The experimental results achieved by ProRank + show its ability to detect protein complexes in protein interaction networks.

View Article: PubMed Central - HTML - PubMed

Affiliation: College of Information Technology, United Arab Emirates University (UAEU, Al Ain 17551, United Arab Emirates. eileen.hanna@uaeu.ac.ae.

ABSTRACT

Background: Developing suitable methods for the identification of protein complexes remains an active research area. It is important since it allows better understanding of cellular functions as well as malfunctions and it consequently leads to producing more effective cures for diseases. In this context, various computational approaches were introduced to complement high-throughput experimental methods which typically involve large datasets, are expensive in terms of time and cost, and are usually subject to spurious interactions.

Results: In this paper, we propose ProRank+, a method which detects protein complexes in protein interaction networks. The presented approach is mainly based on a ranking algorithm which sorts proteins according to their importance in the interaction network, and a merging procedure which refines the detected complexes in terms of their protein members. ProRank + was compared to several state-of-the-art approaches in order to show its effectiveness. It was able to detect more protein complexes with higher quality scores.

Conclusions: The experimental results achieved by ProRank + show its ability to detect protein complexes in protein interaction networks. Eventually, the method could potentially identify previously-undiscovered protein complexes.The datasets and source codes are freely available for academic purposes at http://faculty.uaeu.ac.ae/nzaki/Research.htm.

Show MeSH