Limits...
Linear normalised hash function for clustering gene sequences and identifying reference sequences from multiple sequence alignments.

Helal M, Kong F, Chen SC, Zhou F, Dwyer DE, Potter J, Sintchenko V - Microb Inform Exp (2012)

Bottom Line: However, defining the optimal number of clusters, cluster density and boundaries for sets of potentially related sequences of genes with variable degrees of polymorphism remains a significant challenge.The method was evaluated using sets of closely related (16S rRNA gene sequences of Nocardia species) and highly variable (VP1 genomic region of Enterovirus 71) sequences and outperformed existing unsupervised machine learning clustering methods and dimensionality reduction methods.This method does not require prior knowledge of the number of clusters or the distance between clusters, handles clusters of different sizes and shapes, and scales linearly with the dataset.

View Article: PubMed Central - HTML - PubMed

Affiliation: Sydney Emerging Infections and Biosecurity Institute, Sydney Medical School - Westmead, University of Sydney, Sydney, New South Wales, Australia. vitali.sintchenko@swahs.health.nsw.gov.au.

ABSTRACT

Background: Comparative genomics has put additional demands on the assessment of similarity between sequences and their clustering as means for classification. However, defining the optimal number of clusters, cluster density and boundaries for sets of potentially related sequences of genes with variable degrees of polymorphism remains a significant challenge. The aim of this study was to develop a method that would identify the cluster centroids and the optimal number of clusters for a given sensitivity level and could work equally well for the different sequence datasets.

Results: A novel method that combines the linear mapping hash function and multiple sequence alignment (MSA) was developed. This method takes advantage of the already sorted by similarity sequences from the MSA output, and identifies the optimal number of clusters, clusters cut-offs, and clusters centroids that can represent reference gene vouchers for the different species. The linear mapping hash function can map an already ordered by similarity distance matrix to indices to reveal gaps in the values around which the optimal cut-offs of the different clusters can be identified. The method was evaluated using sets of closely related (16S rRNA gene sequences of Nocardia species) and highly variable (VP1 genomic region of Enterovirus 71) sequences and outperformed existing unsupervised machine learning clustering methods and dimensionality reduction methods. This method does not require prior knowledge of the number of clusters or the distance between clusters, handles clusters of different sizes and shapes, and scales linearly with the dataset.

Conclusions: The combination of MSA with the linear mapping hash function is a computationally efficient way of gene sequence clustering and can be a valuable tool for the assessment of similarity, clustering of different microbial genomes, identifying reference sequences, and for the study of evolution of bacteria and viruses.

No MeSH data available.


Related in: MedlinePlus

The optimal number of clusters for the different hash ranges and different number of indices per cluster for (a) Nocardia 16S rRNA 364 sequences of 80 known species; (b) Nocardia 16S rRNA 97 sequences of 4 known species; (c) EV71 109 sequences of 11 known genogroups/subgenogroups; and (d) EV71 500 VP1 sequences of unknown genogroups/subgenogroups.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC3351711&req=5

Figure 2: The optimal number of clusters for the different hash ranges and different number of indices per cluster for (a) Nocardia 16S rRNA 364 sequences of 80 known species; (b) Nocardia 16S rRNA 97 sequences of 4 known species; (c) EV71 109 sequences of 11 known genogroups/subgenogroups; and (d) EV71 500 VP1 sequences of unknown genogroups/subgenogroups.

Mentions: According to the two sensitivity parameters used (hash range and number of hash codes within one cluster), variable cluster boundaries are shown in Figure 1. These variable boundaries will divide any dataset from a minimum number of clusters up to a maximum number of clusters for each dataset, where neither less nor more divisions can be mathematically feasible. Some highly variable datasets produced a minimum of 16 clusters and a maximum of 29 clusters when a hash range of 4 is used and only one code per cluster, which proves that the distances calculated in the distance matrix control the algorithm output and the minimum and maximum number of clusters will vary from one dataset to another. The number of clusters for both datasets used in this study is plotted in Figure 2 as per the two sensitivity parameters. These variable clustering boundaries can be interpreted equally as abstractness levels (the sensitivity of the clustering in order to decide the optimal number of clusters for the dataset), or as hierarchical levels (encapsulations of sub-clusters within larger clusters).


Linear normalised hash function for clustering gene sequences and identifying reference sequences from multiple sequence alignments.

Helal M, Kong F, Chen SC, Zhou F, Dwyer DE, Potter J, Sintchenko V - Microb Inform Exp (2012)

The optimal number of clusters for the different hash ranges and different number of indices per cluster for (a) Nocardia 16S rRNA 364 sequences of 80 known species; (b) Nocardia 16S rRNA 97 sequences of 4 known species; (c) EV71 109 sequences of 11 known genogroups/subgenogroups; and (d) EV71 500 VP1 sequences of unknown genogroups/subgenogroups.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC3351711&req=5

Figure 2: The optimal number of clusters for the different hash ranges and different number of indices per cluster for (a) Nocardia 16S rRNA 364 sequences of 80 known species; (b) Nocardia 16S rRNA 97 sequences of 4 known species; (c) EV71 109 sequences of 11 known genogroups/subgenogroups; and (d) EV71 500 VP1 sequences of unknown genogroups/subgenogroups.
Mentions: According to the two sensitivity parameters used (hash range and number of hash codes within one cluster), variable cluster boundaries are shown in Figure 1. These variable boundaries will divide any dataset from a minimum number of clusters up to a maximum number of clusters for each dataset, where neither less nor more divisions can be mathematically feasible. Some highly variable datasets produced a minimum of 16 clusters and a maximum of 29 clusters when a hash range of 4 is used and only one code per cluster, which proves that the distances calculated in the distance matrix control the algorithm output and the minimum and maximum number of clusters will vary from one dataset to another. The number of clusters for both datasets used in this study is plotted in Figure 2 as per the two sensitivity parameters. These variable clustering boundaries can be interpreted equally as abstractness levels (the sensitivity of the clustering in order to decide the optimal number of clusters for the dataset), or as hierarchical levels (encapsulations of sub-clusters within larger clusters).

Bottom Line: However, defining the optimal number of clusters, cluster density and boundaries for sets of potentially related sequences of genes with variable degrees of polymorphism remains a significant challenge.The method was evaluated using sets of closely related (16S rRNA gene sequences of Nocardia species) and highly variable (VP1 genomic region of Enterovirus 71) sequences and outperformed existing unsupervised machine learning clustering methods and dimensionality reduction methods.This method does not require prior knowledge of the number of clusters or the distance between clusters, handles clusters of different sizes and shapes, and scales linearly with the dataset.

View Article: PubMed Central - HTML - PubMed

Affiliation: Sydney Emerging Infections and Biosecurity Institute, Sydney Medical School - Westmead, University of Sydney, Sydney, New South Wales, Australia. vitali.sintchenko@swahs.health.nsw.gov.au.

ABSTRACT

Background: Comparative genomics has put additional demands on the assessment of similarity between sequences and their clustering as means for classification. However, defining the optimal number of clusters, cluster density and boundaries for sets of potentially related sequences of genes with variable degrees of polymorphism remains a significant challenge. The aim of this study was to develop a method that would identify the cluster centroids and the optimal number of clusters for a given sensitivity level and could work equally well for the different sequence datasets.

Results: A novel method that combines the linear mapping hash function and multiple sequence alignment (MSA) was developed. This method takes advantage of the already sorted by similarity sequences from the MSA output, and identifies the optimal number of clusters, clusters cut-offs, and clusters centroids that can represent reference gene vouchers for the different species. The linear mapping hash function can map an already ordered by similarity distance matrix to indices to reveal gaps in the values around which the optimal cut-offs of the different clusters can be identified. The method was evaluated using sets of closely related (16S rRNA gene sequences of Nocardia species) and highly variable (VP1 genomic region of Enterovirus 71) sequences and outperformed existing unsupervised machine learning clustering methods and dimensionality reduction methods. This method does not require prior knowledge of the number of clusters or the distance between clusters, handles clusters of different sizes and shapes, and scales linearly with the dataset.

Conclusions: The combination of MSA with the linear mapping hash function is a computationally efficient way of gene sequence clustering and can be a valuable tool for the assessment of similarity, clustering of different microbial genomes, identifying reference sequences, and for the study of evolution of bacteria and viruses.

No MeSH data available.


Related in: MedlinePlus