An improved string composition method for sequence comparison

Abstract

BackgroundHistorically, two categories of computational algorithms (alignment-based and alignment-free) have been applied to sequence comparison–one of the most fundamental issues in bioinformatics. Multiple sequence alignment, although dominantly used by biologists, possesses both fundamental as well as computational limitations. Consequently, alignment-free methods have been explored as important alternatives in estimating sequence similarity. Of the alignment-free methods, the string composition vector (CV) methods, which use the frequencies of nucleotide or amino acid strings to represent sequence information, show promising results in genome sequence comparison of prokaryotes. The existing CV-based methods, however, suffer certain statistical problems, thereby underestimating the amount of evolutionary information in genetic sequences.ResultsWe show that the existing string composition based methods have two problems, one related to the Markov model assumption and the other associated with the denominator of the frequency normalization equation. We propose an improved complete composition vector method under the assumption of a uniform and independent model to estimate sequence information contributing to selection for sequence comparison. Phylogenetic analyses using both simulated and experimental data sets demonstrate that our new method is more robust compared with existing counterparts and comparable in robustness with alignment-based methods.ConclusionWe observed two problems existing in the currently used string composition methods and proposed a new robust method for the estimation of evolutionary information of genetic sequences. In addition, we discussed that it might not be necessary to use relatively long strings to build a complete composition vector (CCV), due to the overlapping nature of vector strings with a variable length. We suggested a practical approach for the choice of an optimal string length to construct the CCV.

Using Markov model to improve word normalization algorithm for biological sequence comparison

“…Lu et al (2008) have found two problems associated with composition vector methods: (a) there is a positive correlation between the observed count c(w k,1 … w k,k ) and the estimated expected count c 0 (w k,1 … w k,k ), and (b) a square root needs to be applied to the denominator. Without such an operation, the normalized count tends to be over-standardized.…”

“…This enables building more complex, biologically realistic models with large numbers of parameters, such as Markov model (Pham and Zuegg 2004;Hao and Qi 2004;Wu et al 2006), mix model such as Markov model plus k-word distributions Kantorovitz et al 2007), and Bernoulli model assuming a known word distribution (Lu et al 2008). Although the more complex models in biological sequence comparison are general improvements over the traditional word-based models (Blaisdell 1986;Wu et al 1997Wu et al , 2001Stuart et al 2002), some problems in developing statistical models and estimating the parameters of the complex models have impeded the development and adoption of these or other more complex models.…”

“…Recently, Lu et al (2008) proposed an improved composition vector (ICV) method that takes into consideration the above problems and achieves better performance in sequence comparison. The word normalization in improved composition vector is desirable, but not sufficient, because much effort of the word normalization aims to find better ways of utilizing evolution information.…”

“…; w k;m Þ; 0 otherwiseWe then obtain where V½ f ðw k ÞjM is the variance of the k-word frequencies under Markov model (M). Similar to the work ofLu et al (2008), we normalized the frequencies of the k-words under Markov model (M), denoted by NF, as follows:…”

Identification and isolation of full-length cDNA sequences by sequencing and analysis of expressed sequence tags from guarana (Paullinia cupana)

“…This trend will probably continue with new transformations that emerge from its integrative, quantitative and impressive nature (Fuchs, 2002). This growing proliferation of data from biological sequences made possible the development of many algorithms for the analysis and mining of knowledge (Lu et al, 2008).…”