Kaohsiung J Med Sci April 2009 • Vol 25 • No 4 165 Single nucleotide polymorphisms (SNPs) are the most
common genetic variations in heredity. An SNP is a var-iation of the DNA sequence caused by the replacement of one nucleotide with another, insertion of one or more nucleotides, or nucleotide deletion. Knowledge
of SNPs constitutes useful information for personalized medicine [1,2]. Using SNPs, a variety of genetic diseases can be diagnosed, and side effects of medication can be prevented, thus increasing therapeutic efficiency [3].
Comparative SNP searches are very useful; there-fore, software packages have been developed to facili-tate comparisons in the literature. SNPper [4] and FESD [5], for example, rely on the chromosome position, the cytogenetic band position and the name to query information. SNPHunter [6] provides SNP screening, se-lection and acquisition. Genewindow provides an in-teractive tool for visualization of genomic variations [7]. Received: Dec 26, 2008 Accepted: Mar 27, 2009
Address correspondence and reprint requests to: Dr Hsueh-Wei Chang, Department of Biomedical Science and Environmental Biology, Kaohsiung Medical University, 100 Shih-Chuan 1st Road,
Kaohsiung 807, Taiwan. E-mail: [email protected]
D
YNAMIC
P
ROGRAMMING FOR
S
INGLE
N
UCLEOTIDE
P
OLYMORPHISM
ID I
DENTIFICATION IN
S
YSTEMATIC
A
SSOCIATION
S
TUDIES
Cheng-Hong Yang,1Li-Yeh Chuang,2Yu-Huei Cheng,1Cheng-Hao Wen,3and Hsueh-Wei Chang3,4,5
1Department of Electronic Engineering, National Kaohsiung University of Applied Sciences; 2Department of Chemical Engineering, I-Shou University; 3Faculty of Biomedical Science and Environmental Biology, 4Graduate Institute of Natural Products, College of Pharmacy, and 5Center of Excellence for Environmental
Medicine, Kaohsiung Medical University, Kaohsiung, Taiwan.
Single nucleotide polymorphisms (SNPs) play an important role in personalized medicine. However, the SNP data reported in many association studies provide only the SNP nucleotide/ amino acid position, without providing the SNP ID recorded in National Center for Biotechnology Information databases. A tool with the ability to provide SNP ID identification, with a user-friendly interface, is needed. In this paper, a dynamic programming algorithm was used to compare homologs when the processed input sequence is aligned with the SNP FASTA database. Our novel system provides a web-based tool that uses the National Center for Biotechnology Information dbSNP database, which provides SNP sequence identification and SNP FASTA formats. Freely selectable sequence formats for alignment can be used, including general sequence formats (ACGT, [dNTP1/dNTP2] or IUPAC formats) and orientation with bidirectional sequence match-ing. In contrast to the National Center for Biotechnology Information SNP-BLAST, the proposed system always provides the correct targeted SNP ID (SNP hit), as well as nearby SNPs (flanking hits), arranged in their chromosomal order and contig positions. The system also solves problems inherent in SNP-BLAST, which cannot always provide the correct SNP ID for a given input sequence. Therefore, this system constitutes a novel application which uses dynamic programming to identify SNP IDs from the literature and keyed-in sequences for systematic association studies. It is freely available at http://bio.kuas.edu.tw/SNPosition/.
Key Words:association studies, BLAST, BLAT, dynamic programming,
single nucleotide polymorphisms (Kaohsiung J Med Sci 2009;25:165–76)
However, these systems are incapable of screening a manually keyed-in sequence for SNP ID identification. Recently, many servers have begun to provide screening functions for SNP sequences; both SNPServer [8] and GeneSNPs [9] provide a Basic Local Alignment Search Tool (BLAST) function for a sequence; how-ever, the BLAST function, by its very nature, produces some problematic results, which will be described later. Similarly, the National Center for Biotechnology Information (NCBI) BLAST [10] showed different results when different sequences were used, such as those from blastn and megablast. The NCBI dbSNP [11] contains a BLAST program for SNPs called SNP-BLAST [12]. SNP-SNP-BLAST contains numerous data-banks that provide the BLAST function, but there are still problems when performing an SNP ID query using manually keyed-in sequence. For instance, when using the blastn function of SNP-BLAST with me-gablast and blastn without meme-gablast to perform the BLAST function for a partial sequence, the results did not always show the original keyed-in rs#. The BLAST results did not provide high scores or high expected values (which means that the specificity is low), or a corresponding SNP ID could not be found (please see details in the Results and Discussion section). There-fore, in order to correctly query the SNP ID, the system described here uses a dynamic programming algorithm to compare homologs of the input sequences with the FASTA sequence database to avoid such problems.
We used a JAVA-based server to solve all of the above problems. This server includes data for human, mouse and rat SNPs. The system uses dynamic pro-gramming [13–15], which allows users to key-in the sequence in different formats, and provides SNP IDs within the input range, as well as allowing users to identify NCBI SNP IDs and user-defined SNPs in an unknown sequence. In addition, a search function for SNP FASTA formats is provided, which overcomes the limitations of the NCBI system, in which SNP-BLAST does not provide only the correct SNP ID.
M
ETHODSSystem design
A dynamic programming algorithm is one of the most widely used paradigms in computational molecular biology; it can evaluate a search space in polynomial time. Dynamic programming algorithms have been
successfully applied to sequence alignment, gene re-cognition, RNA structure prediction, and to tag SNP selection. In this paper, we describe the development of a web-based SNP information platform. The system provides users with an input function for unknown sequences and performs sequence alignments with the rs_FASTA sequence in the databanks. The align-ment results show the SNP ID, contig position, and FASTA sequence information within the input se-quence (a sese-quence in FASTA format begins with a single-line description, followed by lines of sequence data). The system also searches for 5⬘ and 3⬘ flanking sequences (not including alleles) of SNP FASTA se-quences to provide related information about nearby SNPs. Finally, the SNP information is presented graph-ically (see Figures 1D and 1E, and Figure 2).
The flowchart for our system, shown in Figure 3, is divided into four modules: an input module, a se-quence data preprocessing module, a sese-quence align-ment module, and a visualization module. In the input module, two types of input formats, sequence input and ID input, are allowed. For the sequence input for-mats, the users may input non-preprocessed DNA sequences or SNP sequences of alleles. The input se-quence is then transferred into the sese-quence data pre-processing module which deletes unnecessary blanks and filters for non-base symbols. The processed se-quence is then aligned with the FASTA sese-quence data-base in the sequence alignment module. Finally, the alignment results are graphically depicted in the visualization module. The database used by this system contains data for human (ftp://ftp.ncbi.nih. gov/snp/ human/), mouse (ftp://ftp.ncbi.nih.gov/snp/mouse/) and rat (ftp://ftp.ncbi.nih.gov/snp/rat/) SNPs in NCBI dbSNP version b126 [11]. Each module in our system is described in more detail below.
Input module
In the input model, a free sequence format is allowed; DNA sequence formats (A, C, G, or T), dNTP1/dNTP2 sequence formats, and the IUPAC format can be used. The input sequence is processed for SNP ID retrieval in the input orientation. If no SNP is found, the input sequence is automatically transferred to its comple-mentary sequence for re-retrieval.
Sequence data preprocessing module
The purpose of this module is to delete unnecessary blanks and line-feeds, and to filter non-base symbols
other than A, T, C, and G. The IUPAC symbols M, R, W, S, Y, K, V, H, D, B, and N are reserved for the transformation of [dNTP1/dNTP2] to the IUPAC format.
Sequence alignment module
The processed input sequence is aligned with the FASTA sequence database, and a dynamic program-ming method [13–15] is used to compare their ho-mologs. To determine whether the user’s input sequence contains variations or not, the sequence is aligned with the SNP FASTA sequence data. The
system compares the bases of an input data sequence one with the SNP rs_FASTA sequences to identify the sequence with the greatest similarity. First, the SNP FASTA sequence and the input sequence of the suffix edit distance E are calculated. Where P(i) represents the base of the user’s input sequence in index i, i= 1, 2, …, m, and m is the user’s input sequence length; T(j) is the base of SNP FASTA sequence in index j, j= 1, 2, …, n, and n is the SNP FASTA sequence’s length. The procedure for the suffix edit distance is given below. The computational performance when carrying out the proposed analysis is O(Nmn), where
SNP1 SNP2 SNP3 SNP4 SNP5
SNP:
Contig position Visual sequence
SNP1 SNP2 Input sequence SNP3 SNP4 SNP5 SNP: Contig pos: Visual sequence Visual sequence 1138743 1138856
Visual sequence Right interval
sCtgPos eCtgPos
maxCtg – minCtg
(minCtg) (maxCtg)
The maximum SNP contig position The minimum SNP contig position
Left interval SNP1 A B C D E SNP5 alleleCtgPos allelePos sPos Length Input sequence: SNP b. Allele match SNP a. Complete match
SNP c. Part allele match
SNP
SNP d. No allele match
SNP e. Invalidated match
SNP f. Complete flanking match
Flanking
sequence SNP g. Incomplete flanking match
SNP fasta:
Input sequence
Figure 1.Examples and principles for visualization of the output of target single nucleotide polymorphism (SNP) alignment. (A) The related SNPs 1–5 are linearly arranged according to the contig position of the visualized sequence. (B) A graph of the whole sequence based on the interval distance and flanking region of both sides. This range is then reflected to the ruler with a scale, as shown in (E). (C) A related position of the input sequence to the SNP FASTA sequence and the visualized sequence. Only one SNP FASTA sequence is shown although several overlapping SNP FASTA sequences are projected into the visualized sequence. (D) SNP color conditions for SNPs a–g (left side) and the results of the alignment (right side) are displayed individually. The conditions for SNPs a–e and SNPs f and g are discussed with the SNP and flanking region, respectively. Refer to the example with sequences in the text. (E) The alignment results of the input sequence with respect to the visualized sequence. The SNP linear arrangement is based on the contig position.
N is the number of SNPs in NCBI dbSNP, and m and n are as described above.
The procedure for determining the suffix edit distance E: for i 0 to m do E(i, 0) i end for j 0 to n do E(0, j) 0 end for i 1 to m do for j 0 to n do
if(T(j)= P(i)) then
E(i, j) (i − 1, j − 1) else
minMIN[E(i −1, j), E(i, j −1)] E(i, j) min + 1
end end return E
To find the partially homologous sequence for users, the maximum error tolerance for the input se-quence is accepted (Equation 1). Once the error count for aligning P and T is equal to or less than the maxi-mum error tolerance, the input sequence is success-fully aligned with the SNP FASTA sequence.
A
B
Figure 2. The visual output of our system can correctly match single nucleotide polymorphisms (SNP), for example: (A) rs13083205; and (B) rs6167569. The SNP ID rs#, organism, chromosome location, SNP and flanking hits, as well as their chromosome contig positions are displayed. In this system, the relationship between these SNPs is provided in the order of chromosome contig. The SNP hit and flanking hits for the SNP are marked by red and gray colors, respectively.
1. SNP ID
2. Contig position order 3. Color mark 4. Mismatch mark 5. Input sequence 6. Position scale 7. FASTA information 1. Sequence input Input a sequence for human, mouse
or rat Visualization module Input module Sequence data preprocessing module Sequence alignment module 2. ID input Input rs#, ss#, gene name or gene ID dbSNP FASTA sequence db ID rs# Response sequence Hit sequence FASTA sequence Request sequence
Free format sequence System format sequence
The maximum error tolerance number
= (input sequence length) × (tolerant error rate) (Equation 1)
The homologous sequence can be obtained using the previously determined suffix edit distance E and the maximum error tolerance number based on backwards dynamic programming. Once the suffix edit distance E is less than or equal to the maximum error tol-erance number, it is processed. The backward sequence is the homologous sequence that fits with the analog. For example, if an input sequence contains the bases (nucleotides) TAGC, the maximum tolerance error rate is 20%. When the input sequence is aligned with the 10-bases-long SNP FASTA sequence, TGGATACCAT, the maximum error tolerance number is 10× 0.2 = 2. In other words, only two or fewer error alignments are allowed in this case (Figure 4). The boldface arrows in Figure 4 indicate the output of an agreeable homologous alignment, for which the homologous sequences are: (1) TG; (2) TGG; (3) TGGA; and (4) TA.
Visualization module
In the visualization module, the SNPs to be compared and the variations in the SNP flanking regions are pre-sented graphically. The SNPs compared (e.g. SNP1– SNP5) are linearly arranged according to their contig positions, as shown in Figure 1A. The visualization graph of the whole sequence is output based on the interval distance and the flanking regions on both sides, as shown in Figure 1B. Because the input se-quence only shows the sese-quence length, we obtained a similar initial position and allele position for the SNP FASTA sequence using a similar alignment (Figure 1C). In addition, we determined the initial contig position
(sCtgPos) and the final contig position (eCtgPos) of the input sequence by using the allele contig position (allelCtgPos) of the SNP within the whole sequence as shown in Figure 1C. The calculation is shown below (Equations 2 and 3).
sCtgPos=alleleCtgPos−(allelePos−sPos) (Equation 2) eCtgPos= sCtgPos + length (Equation 3) Because the comparison reveals many different results, the proposed system shows the different results coded in different colors, as illustrated in Figure 1D. For alignment with targeting to the NCBI SNP, SNPs a–e are displayed with different colors under different circumstances in Figure 1D. For a flanking sequence alignment without targeting to the NCBI SNP, SNPs f and g are also displayed with different colors. More-over, the matched SNPs or variations in the flanking sequence are displayed individually in the input se-quence, as shown on the right hand side of Figure 1D. Finally, the results of the input sequence are homol-ogous to the SNP linear alignment according to the contig position (Figure 1E). When the visualization is processed, the minimum and maximum contig posi-tions, e.g. 1138743 and 1138856 respectively, must be found within all of the matched SNPs.
R
ESULTS ANDD
ISCUSSIONNinety percent of polymorphisms in the human ge-nome are SNPs. Knowledge of these SNPs is regarded as an important resource for personalized medicine studies. A table containing 172 different sequences from 23 different chromosomes is provided online at http:// bio.kuas.edu.tw/dynamicSNP/ncbi-compare.jsp, and the results obtained by SNP-BLAST in NCBI and our proposed system are compared. Here, the input sequence containing Homo sapiens rs13083205 [A/C] for SNP-BLAST of NCBI is described below. As an example, we choose an input sequence copied from an SNP ID—namely for rs13083205 SNP ID [Homo sapiens]: AAAATTAATATTCTTCAAAATTMTTTCTT CTAAAT (the length of the input sequence is shown in blue as indicated by arrows in Figure 2A; M indi-cates the SNP site).
When the blastn with megablast (default in the NCBI) SNP-BLAST programs are used, these se-quences gave different results for rs13083205. When
T G G A T A C C A T 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 0 1 1 1 1 0 T 1 1 2 2 1 1 0 1 2 1 1 A 2 2 1 2 2 2 1 2 3 2 2 G 3 3 2 1 2 3 2 3 4 3 3 G 4 (1) (2) (3) (4)
Figure 4. Homologous alignment and possible homologous sequences.
Figure 5. Single nucleotide polymorphism (SNP) ID identification results obtained using the National Center for Biotechnology Information SNP-tool. The results obtained with SNP-BLAST are separated, showing blastn with megablast on the left side and blastn without megablast on the right side. SNP ID identification result of input sequence with SNP in: (A) International Union of Pure and Applied Chemistry format “M”; (B) [A/C]; (C) A; and (D) C.
the sequence was input in the IUPAC format, i.e. AAA-ATTAATATTCTTCAAAATTMTTTCTTCTAAAT (Figure 5A) and [dNTP1/dNTP2] sequence format, i.e. AAAATTAATATTCTTCAAAATT[A/C]TTTC TTCTA-AAT (Figure 5B), SNP-BLAST was unable to locate a significantly similar alignment, i.e. there is no matched SNP. Because the alleles of SNP ID rs13083205 are A/C, the input sequence is a general sequence con-taining the allele base “A” (Figure 5C); the results of SNP-BLAST were rs13083390, rs13063760, rs12495488, rs4452373, rs2056549, rs1464651, rs1464650, rs1464649, and rs13083205. The “scores” were all 65.8 and the “expected” value was 2e–09, except for rs13083205 (score= 62.1 and expected = 3e–08). Most of these SNP IDs provided in the BLAST results were unable to provide the correct SNP ID from the input sequence, i.e. the input rs13083205 SNP ID-containing sequence. Furthermore, the rs13083205 was hit with low per-formance (low score bits and high expected values) compared with the other hits. On the other hand, if the input sequence is a general sequence containing allele base “C” (Figure 5D), the results of SNP-BLAST show “no significant similarity found”. Therefore, the search in SNP-BLAST is sequence format-dependent and differs on a case-by-case basis. In this case, SNP-BLAST does not reveal a single positive result for the C allele of rs13083205, whereas under other circum-stances, no significant hit is retrieved by SNP-BLAST. In some cases, SNP IDs are hit without including the target SNP ID. These are common problems for SNP-BLAST in the NCBI when the SNP-BLAST function is per-formed by inputting a sequence to identify SNP ID or IDs within a sequence. These results can mislead researchers and prevent them from obtaining the correctly matched SNP ID in a trial.
Moreover, the input sequence produces different results when using the SNP-BLAST program blastn of SNP-BLAST in the NCBI without megablast. When the sequence is input in the IUPAC format “M” (Figure 5E), the BLAST results included a total of nine SNP IDs: rs13083390, rs13083205, rs13063760, rs12495488, rs4452373, rs2056549, rs1464651, rs1464650 and rs1464649 with a score of 67.2 and an expected value of 2e–09. However, all of the eleven SNPs in rs13083205 share the highest score; therefore, NCBI SNP-BLAST cannot provide the correct SNP ID(s) for this input se-quence. Similarly, when the input sequence is entered in the SNP nucleotide format [A/C] (Figure 5F) or A (Figure 5G), nine matched SNPs are retrieved by SNP
BLAST: rs13083390, rs13063760, rs12495488, rs4452373, rs2056549, rs1464651, rs1464650, rs1464649 and
rs13083205 with a score of 55.4–57.2 or 60.8–64.4 and
an expected value of 4e–08 to 4e–09 or 2e–10 to 4e–09, respectively. In this case, however, the matched rs13083205 did not have the highest score; thus users are unable to determine whether it is the correct SNP ID. For the sequence AAAATTAATATTCTTCAAAAT-TCTTTCTTCTAAAT (Figure 5H), nine matched SNPs were retrieved by SNP BLAST: rs13083205, rs13083390, rs13063760, rs12495488, rs4452373, rs2056549, rs1464651, rs1464650, rs1464649, rs57392751, and rs12357542, with a score of 44.6–60.8 and an expected value of 0.003 to 4e–08. In this case, rs13083205 showed the highest score with SNP-BLAST. However, different kinds of sequence formats may have differ-ent performance for SNP ID retrieval in the NCBI SNP-BLAST tool. In general, this will cause confusion, and users cannot precisely identify the target SNP ID for a given input sequence.
The NCBI SNP-BLAST tool shows similar problems for other types of sequences from other species, such as Mus musculus. For example, the rs6167569 sequence is TCTTGCGTAGATCCGTCACAGCCCT[C/T]TTTC-ACCCGCCAGGGCTCCCGACAA. When using blastn with megablast, the allele of rs6167569 with T pro-vided the correct SNP IDs in SNP-BLAST with a score 93.5 and an expected value of 7e–18, but the allele of rs6167569 with Y, C and [C/T] did not match the cor-rect SNP ID. Using blastn without megablast, rs6167569 contains different sequences, including [C/T], Y, C and T. They all provided the same BLAST results with three SNP IDs: rs6167569, rs6167516 and rs6167013; of these, rs6167569 attained the highest scores. Therefore, NCBI SNP-BLAST may provide too much data to de-termine a correct SNP ID for the input sequence, or it cannot provide the SNP ID at all. In addition, the SNPs and flanking hits are not marked.
In our novel system, we used the same sequences for rs13083205, located in human chromosome 3, as for SNP-BLAST. The sequence contains SNPs in the nucleotide format R, [A/G], A or G, and the choice for mismatched bases is 1, by default. The four sequence inputs show the same results, and all are correctly matched to the SNP sequence rs13083205, as shown in Figure 2A. Similarly, with rs6167569, located in mouse chromosome 15, the different sequences including [C/T], Y, C and T, were all matched correctly to only a single SNP, rs6167569, as shown in Figure 2B. In
addition, the nearby SNP IDs rs6167013 and rs6167516 are provided as flanking hits distinguished from the SNP hit itself. This provides a complete topography of the input sequence, and it corrects the inherent problem of the NCBI SNP-BLAST tool. In Figure 2A and 2B, the SNP contig position is presented graphi-cally. The color coding of the visualization helps the biologists to identify all of the relevant data contained in the SNP sequence. In addition, users can test their SNPs for novelty if they are unable to find the corre-sponding SNP ID in our proposed system, i.e. user-defined SNPs marked with a bar were discovered.
We designed an experiment to test how often the SNP ID within the input sequence is incorrectly pro-vided using the method shown in Figure 5. Using 172 different sequences from 23 different chromosomes (a list is provided online at http://bio.kuas.edu.tw/ dynamicSNP/ncbi-compare.jsp), we found that our system achieved a success rate of 100%, while the NCBI SNP-BLAST tool, with and without the megablast func-tion, provided the complete score matched (similar to Figure 5H; correct SNP ID hit is the single highest score among hits) for 13.34% versus 40.70%, the high score matched (Figure 5E; several hits share the high-est score but the correct SNP ID hit is not shown in the first row) for 11.05% versus 34.88%, no high score matched (Figures 5C, 5F and 5G; the correct SNP ID hit is not the highest score) for 6.98% versus 20.35%, and no matched or no significance (similar to Figures
5A, 5B and 5D; no hits provided) for 68.60% versus 4.07%, respectively. These results are provided online at the website address given above. For complete-score-matched, high-score-matched and no-high-score-matched conditions, the system is able to provide the correct SNP ID hits, although it is difficult for most users to identify them. In contrast, no-matched and no-significance situations show false-negative results for the SNP ID hit.
The NCBI BLAST tool uses a heuristic algorithm that seeks local, as opposed to global, alignments and is therefore able to detect relationships among se-quences that share only isolated regions of similarity [16]. It obtains all statistically significant alignments. However, it does not explore the entire search space between two sequences. This is the reason why BLAST sometimes cannot hit the correct SNP ID. The overlap-ping nature of several SNP flanking sequences found by the NCBI SNP-BLAST tool is shown in Figure 6 and in our previous results with SNP ID-info soft-ware [17]. This may partly explain why false-positive hits (providing many hits rather than the correct SNP ID hit alone) for NCBI SNP-BLAST were common in our 172 tests with different sequences from 23 dif-ferent chromosomes. Many SNP flanking sequences commonly overlap each other, although the SNPs are located separately.
Recently, the University of California Santa Cruz Genome Browser Database [18] was developed to
rs17886079 rs34097279 rs1421314 rs17551157 rs17883184 rs17883670 rs35515019 rs17882910 rs17551150 2000 1800 1600 1400 1200 1000 800 600 400 200 0 rs17886079 rs34097279 rs17551157 rs35515019 rs1421314 rs17883184 rs17882910 rs17883670 rs17551150 Input sequence Symbol SNP flanking sequence SNP location A B 1 1
Figure 6. Example of the overlapping nature of several sequences containing single nucleotide polymorphisms. (A) Neighboring environmental viewing using Genewindow software. Nine single nucleotide polymorphisms located in chromosome 17 were selected as an example. The relative locus of the chromosomal contig position is provided. (B) The overlapping relationship between the nine single nucleotide polymorphisms flanking sequences can be seen in (A).
T
able.
Comparison of our novel system and the National Center for Biotechnology Information sear
ch tool SNP-BLAST Information for SNP ID r etrieval NCBI SNP-BLAST Pr oposed system Method BLAST Dynamic pr ogramming/BLAST*/BLA
T* (*is used for the primary sear
ch for
the chr
omosome location to impr
ove the subsequent SNP
identification). Sequence sour ce Only SNP FAST A sequences SNP FAST A sequences ar
e integrated with the chr
omosome and
contig positions.
Accuracy
For input sequence containing NCBI SNP:
For input sequence containing NCBI SNP:
1. Indistinguishable to the SNP
hit and
1. Pr
ovides the corr
ect tar
get SNP
ID (or IDs) for input sequence (SNP
hit).
flanking hits.
2. Pr
ovides the flanking SNP
IDs, i.e. SNP
nearby but not within the
2. Sometimes, no significant hit for SNP
ID.
input sequence (flanking hits).
3. No envir
onment for SNP
ID map, i.e. without
For input sequence without NCBI SNP within:
nearby SNPs map in sequence.
1. Pr
ovides SNPs with flanking hits if no SNP
hit is pr
esent within
For input sequence without NCBI SNP within:
the input sequence.
1. Cannot confirm whether any SNP
IDs exist
2. Pr
esents the position and range of the input sequence under the SNP
within the input sequence.
FAST
A
sequence and envir
onment by visualization.
Positional information
1. Not shown immediately or systemically
.
1. Pr
ovided immediately and systemically
.
(chr
omosome/contig
2. SNP
IDs ar
e not listed in the or
der of 2. All r etrieved SNP IDs ar e displayed in the or der of chr omosome and positions) chr
omosome and contig positions.
contig positions.
Analytic parameters
1. Length and SNP
density of the input
1. Length of input sequence: Retrieval SNP
number is incr
easing with the
sequence ar
e pr
oportional to SNP
number
length of input sequence.
retrieved fr om input sequence. 2. No megablast choice pr oblem. Consistent r esults for SNP ID 2. Megablast: dif fer ent BLAST r esults for SNP-identification.
BLAST with or without megablast function.
3. SNPs listed in the or
der of chr
omosome and contig positions.
3. Listed accor
ding to or
der of scor
e and E values.
a. Highest scor
e and lowest E values ar
e not always r epr esentative corr ect SNP ID. b. Pr
ovides too many SNP
IDs with same
performance (scor
e and E values) to call.
c. The scor
e and E values of SNP-BLAST make
no sense for the SNP
ID identification of an input sequences. Display type 1. SNP ID panel listed. 1. V
isualization for the range and position of the input sequence under SNP
2. Listed in the or der of scor e and E values. ID envir onment. 3. SNP
and flanking hits not marked.
2. Color
-match pr
ovides illustration of SNP
and flanking hits.
3. Only the corr
ect SNP
hit is pr
ovided. Nearby SNPs, which ar
e not included
in the input sequence, ar
e pr
ovided and marked as flanking hits.
Major featur
e
Sear
ches similar flanking sequences for SNP
ID.
1. Identifies SNP
ID within an input sequence.
2. Pr
ovides SNP
hit and flanking hits in the or
der of chr
omosome
and contig positions.
SNP
=
single nucleotide polymorphism; ID
=
SNP
refer
ence cluster identification car
d number; NCBI
=
National Center for Biotechnology Information; BLAST
= Basic Local Alignment Sear ch T ool; BLA T = BLAST -Like Alignment T ool; E = expected value.
provide sequence and annotation data for the genomes of many species and model organisms. With its BLAST-Like Alignment Tool (BLAT) function [19], it provides genome annotations including SNP IDs within the input sequence that are similar to our novel system. However, SNPs and flanking hits are not provided by BLAT. This problem is similar to that of SNP-BLAST in terms of the acceptable sequence formats. The results are sometimes sequence format-dependent and many browser hits are provided for selection. The IUPAC code and [dNTP1/dNTP2] formats for SNP are not suitable for BLAT because this sometimes provides the wrong chromosome location and many browsers with different chromosome locations. Furthermore, SNP-BLAST and BLAT do not indicate mutations other than the SNPs in the NCBI dbSNP. In contrast, our proposed system marks both user-defined SNPs and SNPs in dbSNP. However, because SNP-BLAST and BLAT provide fast SNP retrieval, we introduced these functions into our proposed system for sequences with unknown chromosome locations to improve the screening speed. By using dynamic programming in our system, we overcame the limitations inherent in SNP-BLAST and BLAT, and free formats of SNP input sequences are accepted.
Dynamic programming has been used for sequence alignment [13–15], and it has recently been applied to haplotype block studies [20,21]. However, none of these studies performed SNP ID targeting for input sequences. To our knowledge, we are the first to iden-tify SNP IDs in NCBI rs# using dynamic programming. A general comparison of the performance between NCBI SNP-BLAST and our proposed system is shown in the Table. The results indicate that our system is more precise and more informative for SNP ID iden-tification than the NCBI SNP-BLAST, particularly re-garding SNP ID targeting and general information of the SNP. Flanking hits are provided in the order of the chromosome and contig positions. Once users obtain an SNP ID from NCBI dbSNP, software, including our newly developed SNP-RFLPing [22], can provide SNP information for genotyping in association studies.
C
ONCLUSIONThe NCBI BLAST is currently the most commonly used sequence alignment tool. It continues the heuris-tic calculation concept and requires a shorter time to
search for alignment similarities. Nevertheless, it has limitations, such as its inherent problem in the BLAST program, which uses megablast, and the fact that it cannot always correctly align an input sequence. In this paper, we described the development of a web-based SNP system that uses a dynamic programming algorithm to compare homologs. This system provides the correct SNP ID when entering a sequence, and the reference cluster ID “rs”, the NCBI assay ID, the “ss”, the gene name and the gene ID, which facilitates faster searching of SNP FASTA information. This not only solves the inherent limitations of NCBI SNP-BLAST, but also assists biologists trying to establish a systematic database while conducting association research.
A
CKNOWLEDGMENTSThis work was partly supported by the National Science Council in Taiwan under grants 96-2221-E-214-050-MY3 and 97-2622-E-151-008-CC2, and by the grants KMU-EM-97-1.1a and KMU-EM-98-1.4.
R
EFERENCES1. Suh Y, Vijg J. SNP discovery in associating genetic varia-tion with human disease phenotypes. Mutat Res 2005; 573:41–53.
2. Erichsen HC, Chanock SJ. SNPs in cancer research and treatment. Br J Cancer 2004;90:747–51.
3. Shastry BS. SNPs and haplotypes: genetic markers for disease and drug response (review). Int J Mol Med 2003; 11:379–82.
4. Riva A, Kohane IS. SNPper: retrieval and analysis of human SNPs. Bioinformatics 2002;18:1681–5.
5. Kang HJ, Choi KO, Kim BD, et al. FESD: a Functional Element SNPs Database in humans. Nucleic Acids Res 2005;33:D518–22.
6. Wang L, Liu S, Niu T, et al. SNPHunter: a bioinformatic software for single nucleotide polymorphism data acqui-sition and management. BMC Bioinformatics 2005;6:60. 7. Staats B, Qi L, Beerman M, et al. Genewindow: an
interactive tool for visualization of genomic variation. Nat Genet 2005;37:109–10.
8. Savage D, Batley J, Erwin T, et al. SNPServer: a real-time SNP discovery tool. Nucleic Acids Res 2005;33: W493–5.
9. University of Utah Genome Center. GeneSNPs-blast function. Available at: http://www.genome.utah.edu/ genesnps/cgi-bin/blast.cgi [Date accessed: August 23, 2006]
10. McGinnis S, Madden TL. BLAST: at the core of a pow-erful and diverse set of sequence analysis tools. Nucleic Acids Res 2004;32:W20–5.
11. Sherry ST, Ward MH, Kholodov M, et al. dbSNP: the NCBI database of genetic variation. Nucleic Acids Res 2001;29:308–11.
12. National Center for Biotechnology Information. SNP-BLAST. Available at: http://www.ncbi.nlm.nih.gov/ SNP/snp_blastByOrg.cgi [Date accessed: January 9, 2009]
13. Eddy SR. What is dynamic programming? Nat Biotechnol 2004;22:909–10.
14. Giegerich R. A systematic approach to dynamic pro-gramming in bioinformatics. Bioinformatics 2000;16: 665–77.
15. Needleman SB, Wunsch CD. A general method appli-cable to the search for similarities in the amino acid sequence of two proteins. J Mol Biol 1970;48:443–53. 16. Altschul SF, Madden TL, Schaffer AA, et al. Gapped
BLAST and PSI-BLAST: a new generation of protein
database search programs. Nucleic Acids Res 1997;25: 3389–402.
17. Yang CH, Chuang LY, Cheng YH, et al. SNP ID-info: SNP ID searching and visualization platform. OMICS 2008;12:217–26.
18. Hinrichs AS, Karolchik D, Baertsch R, et al. The UCSC Genome Browser Database: update 2006. Nucleic Acids Res 2006;34:D590–8.
19. Kent WJ. BLAT—the BLAST-like alignment tool. Genome Res 2002;12:656–64.
20. Zhang K, Qin Z, Chen T, et al. HapBlock: haplotype block partitioning and tag SNP selection software using a set of dynamic programming algorithms. Bioinformatics 2005;21:131–4.
21. Su SC, Kuo CC, Chen T. Inference of missing SNPs and information quantity measurements for haplotype blocks. Bioinformatics 2005;21:2001–7.
22. Chang HW, Yang CH, Chang PL, et al. SNP-RFLPing: restriction enzyme mining for SNPs in genomes. BMC Genomics 2006;7:30.
在系統相關性研究中使用動態規劃法找尋單核苷酸
多型性身分證 (SNP ID)
楊正宏1 莊麗月2 鄭煜輝1 溫政浩3 張學偉3,4,5 1 高雄應用科技大學 電子工程學系 2 義守大學 化學工程學系 高雄醫學大學 3生物醫學暨環境生物學系 4天然藥物研究所 5環境醫學頂尖研究中心單一核苷酸多型性 (single nucleotide polymorphisms、SNP) 在個人化醫學上扮演 很重要的角色。然而，很多關聯性研究對於 SNP 資訊的描述僅提供 SNP 核苷酸或胺 基酸的位置，而沒有提供 NCBI SNP ID。因此，一個能夠提供 SNP ID 辨識且具有 友善使用者介面的工具是必需的。本研究使用動態規劃方法來比對輸入序列與 SNP
FASTA 資料庫間的相似性，並且提出一個基於 NCBI dbSNP 資料庫的 web-based
系統，來提供 SNP 序列辨識和 SNP FASTA 格式。此系統能夠輸入各種格式的序列， 包括一般化的序列格式 ( 如 ACGT、[dNTP1/dNTP2] 或 IUPAC 格式 ) 和不同方向 的序列。相較於 NCBI SNP-BLAST，本研究提出的系統能依照染色體和 contig 位置 提供正確的目標的 SNP ID (SNP hit)，而且也能提供鄰近的 SNPs (flanking hits)， 並且也解決了 SNP-BLAST 對於輸入序列常常不能夠提供正確 SNP ID 的問題。因 此，我們將動態規劃方法新穎地應用在 SNP ID 搜尋與辨識，讓使用者可以查詢文獻 上的序列及輸入序列中所包含的 SNP ID，而提供系統性的相關性研究。這個程式在 網頁 http://bio.kuas.edu.tw/SNPosition/ 可以免費使用。 關鍵詞：關聯性研究，BLAST，BLAT，動態規劃方法，單一核苷酸多型性 ( 高雄醫誌 2009;25:165–76) 收文日期：97 年 12 月 26 日 接受刊載：98 年 3 月 27 日 通訊作者：張學偉副教授 高雄醫學大學生物醫學暨環境生物學系 高雄市三民區十全一路 100 號
