Limits...
Alignment-free clustering of transcription factor binding motifs using a genetic-k-medoids approach.

Broin PÓ, Smith TJ, Golden AA - BMC Bioinformatics (2015)

Bottom Line: We demonstrate the performance of our algorithm, GMACS, on multiple benchmark motif datasets, comparing results obtained with current leading approaches.Finally, we show the robustness of our algorithm to noise on multiple synthetic datasets consisting of known motifs convolved with varying degrees of noise.Our proposed algorithm is generally applicable to any DNA or protein motifs, can produce highly stable and biologically meaningful clusters, and, by avoiding the problem of frozen sub-alignments, can provide improved results when compared with existing techniques on benchmark datasets.

View Article: PubMed Central - PubMed

Affiliation: Department of Genetics, Albert Einstein College of Medicine, 1301 Morris Park Avenue, Bronx, New York, 10461, USA. pilib.obroin@einstein.yu.edu.

ABSTRACT

Background: Familial binding profiles (FBPs) represent the average binding specificity for a group of structurally related DNA-binding proteins. The construction of such profiles allows the classification of novel motifs based on similarity to known families, can help to reduce redundancy in motif databases and de novo prediction algorithms, and can provide valuable insights into the evolution of binding sites. Many current approaches to automated motif clustering rely on progressive tree-based techniques, and can suffer from so-called frozen sub-alignments, where motifs which are clustered early on in the process remain 'locked' in place despite the potential for better placement at a later stage. In order to avoid this scenario, we have developed a genetic-k-medoids approach which allows motifs to move freely between clusters at any point in the clustering process.

Results: We demonstrate the performance of our algorithm, GMACS, on multiple benchmark motif datasets, comparing results obtained with current leading approaches. The first dataset includes 355 position weight matrices from the TRANSFAC database and indicates that the k-mer frequency vector approach used in GMACS outperforms other motif comparison techniques. We then cluster a set of 79 motifs from the JASPAR database previously used in several motif clustering studies and demonstrate that GMACS can produce a higher number of structurally homogeneous clusters than other methods without the need for a large number of singletons. Finally, we show the robustness of our algorithm to noise on multiple synthetic datasets consisting of known motifs convolved with varying degrees of noise.

Conclusions: Our proposed algorithm is generally applicable to any DNA or protein motifs, can produce highly stable and biologically meaningful clusters, and, by avoiding the problem of frozen sub-alignments, can provide improved results when compared with existing techniques on benchmark datasets.

Show MeSH
Genetic algorithm overview. After the population is initialized, the algorithm iterates between evaluating the solutions from the current population and generating new solutions from selected candidates through various evolutionary operators. Termination usually occurs after a specified number of iterations (‘generations’) have elapsed, when an acceptable solution (such as within an allowable error threshold) has been found, or, after a fixed period during which the fitness has remained relatively constant.
© Copyright Policy - open-access
Related In: Results  -  Collection

License 1 - License 2
getmorefigures.php?uid=PMC4384390&req=5

Fig2: Genetic algorithm overview. After the population is initialized, the algorithm iterates between evaluating the solutions from the current population and generating new solutions from selected candidates through various evolutionary operators. Termination usually occurs after a specified number of iterations (‘generations’) have elapsed, when an acceptable solution (such as within an allowable error threshold) has been found, or, after a fixed period during which the fitness has remained relatively constant.

Mentions: Genetic algorithms, based on early work by Fraser [20] and later popularized by Holland [21] and Goldberg [22], are a stochastic optimization technique making use of a population of candidate solutions. These candidate solutions, commonly encoded as binary strings (although representation as integers and floating-point numbers are also popular), are iteratively evaluated for their effectiveness, or ‘fitness’, for a given problem domain, which is often termed a ‘fitness landscape’ or ‘search space’. The solutions and then combined through the use of evolutionarily inspired genetic operators such as selection, mutation, and crossover, to form the next generation of candidates (Figure 2), with each successive generation providing increasingly ‘fit’ solutions. The parallel search capabilities of GAs (simultaneous sampling of multiple points in the fitness landscape) coupled with their ability to potentially ‘escape’ local minima through the introduction of ‘novelty’ via mutation, make them ideally suited to complex, noisy problem domains, and their use for both multiple sequence alignment [23,24] and motif discovery [25,26] has been well established. By embedding the k-medoids algorithm within a GA framework and choosing a suitable ‘fitness function’ to evaluate candidates, we can leverage the local search capabilities of the k-medoids algorithm while using the GA to both perform global search for the optimal number of clusters, and to provide multiple initializations for the k-medoids algorithm, reducing the potential impact of poorly chosen starting medoids. Below, we detail the implementation and performance of such an alignment-free, non-tree-based approach to the motif clustering problem based on a software tool we have developed called GMACS.Figure 2


Alignment-free clustering of transcription factor binding motifs using a genetic-k-medoids approach.

Broin PÓ, Smith TJ, Golden AA - BMC Bioinformatics (2015)

Genetic algorithm overview. After the population is initialized, the algorithm iterates between evaluating the solutions from the current population and generating new solutions from selected candidates through various evolutionary operators. Termination usually occurs after a specified number of iterations (‘generations’) have elapsed, when an acceptable solution (such as within an allowable error threshold) has been found, or, after a fixed period during which the fitness has remained relatively constant.
© Copyright Policy - open-access
Related In: Results  -  Collection

License 1 - License 2
Show All Figures
getmorefigures.php?uid=PMC4384390&req=5

Fig2: Genetic algorithm overview. After the population is initialized, the algorithm iterates between evaluating the solutions from the current population and generating new solutions from selected candidates through various evolutionary operators. Termination usually occurs after a specified number of iterations (‘generations’) have elapsed, when an acceptable solution (such as within an allowable error threshold) has been found, or, after a fixed period during which the fitness has remained relatively constant.
Mentions: Genetic algorithms, based on early work by Fraser [20] and later popularized by Holland [21] and Goldberg [22], are a stochastic optimization technique making use of a population of candidate solutions. These candidate solutions, commonly encoded as binary strings (although representation as integers and floating-point numbers are also popular), are iteratively evaluated for their effectiveness, or ‘fitness’, for a given problem domain, which is often termed a ‘fitness landscape’ or ‘search space’. The solutions and then combined through the use of evolutionarily inspired genetic operators such as selection, mutation, and crossover, to form the next generation of candidates (Figure 2), with each successive generation providing increasingly ‘fit’ solutions. The parallel search capabilities of GAs (simultaneous sampling of multiple points in the fitness landscape) coupled with their ability to potentially ‘escape’ local minima through the introduction of ‘novelty’ via mutation, make them ideally suited to complex, noisy problem domains, and their use for both multiple sequence alignment [23,24] and motif discovery [25,26] has been well established. By embedding the k-medoids algorithm within a GA framework and choosing a suitable ‘fitness function’ to evaluate candidates, we can leverage the local search capabilities of the k-medoids algorithm while using the GA to both perform global search for the optimal number of clusters, and to provide multiple initializations for the k-medoids algorithm, reducing the potential impact of poorly chosen starting medoids. Below, we detail the implementation and performance of such an alignment-free, non-tree-based approach to the motif clustering problem based on a software tool we have developed called GMACS.Figure 2

Bottom Line: We demonstrate the performance of our algorithm, GMACS, on multiple benchmark motif datasets, comparing results obtained with current leading approaches.Finally, we show the robustness of our algorithm to noise on multiple synthetic datasets consisting of known motifs convolved with varying degrees of noise.Our proposed algorithm is generally applicable to any DNA or protein motifs, can produce highly stable and biologically meaningful clusters, and, by avoiding the problem of frozen sub-alignments, can provide improved results when compared with existing techniques on benchmark datasets.

View Article: PubMed Central - PubMed

Affiliation: Department of Genetics, Albert Einstein College of Medicine, 1301 Morris Park Avenue, Bronx, New York, 10461, USA. pilib.obroin@einstein.yu.edu.

ABSTRACT

Background: Familial binding profiles (FBPs) represent the average binding specificity for a group of structurally related DNA-binding proteins. The construction of such profiles allows the classification of novel motifs based on similarity to known families, can help to reduce redundancy in motif databases and de novo prediction algorithms, and can provide valuable insights into the evolution of binding sites. Many current approaches to automated motif clustering rely on progressive tree-based techniques, and can suffer from so-called frozen sub-alignments, where motifs which are clustered early on in the process remain 'locked' in place despite the potential for better placement at a later stage. In order to avoid this scenario, we have developed a genetic-k-medoids approach which allows motifs to move freely between clusters at any point in the clustering process.

Results: We demonstrate the performance of our algorithm, GMACS, on multiple benchmark motif datasets, comparing results obtained with current leading approaches. The first dataset includes 355 position weight matrices from the TRANSFAC database and indicates that the k-mer frequency vector approach used in GMACS outperforms other motif comparison techniques. We then cluster a set of 79 motifs from the JASPAR database previously used in several motif clustering studies and demonstrate that GMACS can produce a higher number of structurally homogeneous clusters than other methods without the need for a large number of singletons. Finally, we show the robustness of our algorithm to noise on multiple synthetic datasets consisting of known motifs convolved with varying degrees of noise.

Conclusions: Our proposed algorithm is generally applicable to any DNA or protein motifs, can produce highly stable and biologically meaningful clusters, and, by avoiding the problem of frozen sub-alignments, can provide improved results when compared with existing techniques on benchmark datasets.

Show MeSH