Chapter 5:  The Single Nucleotide Polymorphism Database (dbSNP) of Nucleotide Sequence Variation

Physical Mapping

In the physical mapping of nucleotide sequences, variations are used as positional markers. When mapped to a unique location in a genome, variation markers work with the same logic as Sequence Tagged Sites (STSs) or framework microsatellite markers. As is the case for STSs, the position of a variation is defined by its unique flanking sequence, and hence, variations can serve as stable landmarks in the genome, even if the variation is fixed for one allele in a sample. When multiple alleles are observed in a sample pedigree, pedigree members can be tested for variation genotypes as in traditional physical mapping studies.

Functional Analysis

Variations that occur in functional regions of genes or in conserved non-coding regions might cause significant changes in the complement of transcribed sequences. This can lead to changes in protein expression that can affect aspects of the phenotype such as metabolism or cell signaling. We note possible functional implications of DNA sequence variations in dbSNP in terms of how the variation alters mRNA transcripts.

Association Studies

The associations between variations and complex genetic traits are more ambiguous than simple, single-gene mutations that lead to a phenotypic change. When multiple genes are involved in a trait, then the identification of the genetic causes of the trait requires the identification of the chromosomal segment combinations, or haplotypes, that carry the putative gene variants.

Evolutionary Studies

The variations in dbSNP currently represent an uneven but large sampling of genome diversity. The human data in dbSNP include submissions from the SNP Consortium, variations mined from genome sequence as part of the human genome project, and individual lab contributions of variations in specific genes, mRNAs, ESTs, or genomic regions.

Null Results Are Important

Systematic surveys of sequence variation will undoubtedly reveal sequences that are invariant in the sample. These observations can be submitted to dbSNP as NoVariation records that record the sequence, the population, and the sample size that were used in the survey.

Entrez SNP

dbSNP is a part of the Entrez integrated information retrieval system (Chapter 15) and may be searched using either qualifiers (aliases) or a combination of 25 different search fields. A complete list of the qualifiers and search fields can be found on the Entrez SNP site.

Single Record (Search by ID Number) Query in dbSNP

Use this query module to select SNPs based on dbSNP record identifiers. These include reference SNP (refSNP) cluster ID numbers (rs#), submitted SNP Accession numbers (ss#), local (or submitter) IDs, Celera IDs, Genbank accession numbers, and STS accession numbers.

SNP Submission Information Queries

Table 1. Method classes organize submissions by a (more...)

Table 1. Method classes organize submissions by a general methodological or experimental approach to assaying for variation in the DNA sequence.

Method class	Class code in Sybase, ASN.1, and XML
Denaturing high pressure liquid chromatography (DHPLC)	1
DNA hybridization	2
Computational analysis	3
Single-stranded conformational polymorphism (SSCP)	5
Other	6
Unknown	7
Restriction fragment length polymorphism (RFLP)	8
Direct DNA sequencing	9

Table 2. Population classes organize population samples (more...)

Table 2. Population classes organize population samples by geographic region.

Population class	Description	Population class in Sybase, ASN.1, and XML
Central Asia	Samples from Russia and its satellite Republics and from nations bordering the Indian Ocean between East Asia and the Persian Gulf regions.	8
Central/South Africa	Samples from nations south of the Equator, Madagascar, and neighboring island nations.	4
Central/South America	Samples from Mainland Central and South America and island nations of the western Atlantic, Gulf of Mexico, and Eastern Pacific.	10
East Asia	Samples from eastern and south eastern Mainland Asia and from Northern Pacific island nations.	6
Europe	Samples from Europe north and west of Caucasus Mountains, Scandinavia, and Atlantic islands.	5
Multi-National	Samples that were designated to maximize measures of heterogeneity or sample human diversity in a global fashion. Examples include OEFNER|GLOBAL and the CEPH repository.	1
North America	All samples north of the Tropic of Cancer, including defined samples of United States Caucasians, African Americans, Hispanic Americans, and the NHGRI polymorphism discovery resource (NCBI|NIHPDR).	9
North/East Africa and Middle East	Samples collected from North Africa (including the Sahara desert), East Africa (south to the Equator), Levant, and the Persian Gulf.	2
Pacific	Samples from Australia, New Zealand, Central and Southern Pacific Islands, and Southeast Asian peninsular/island nations.	7
Unknown	Samples with unknown geographic provinces that are not global in nature.	11
West Africa	Sub-Saharan nations bordering the Atlantic north of the Congo River and central/southern Atlantic island nations.	3

dbSNP Batch Query

Use sets of variation IDs (including RefSNP (rs) IDs, Submitted SNP (ss) IDs, and Local SNP IDs) collected from other queries to generate a variety of SNP reports.

Locus Information Query

This search was originally accomplished by LocusLink, which has now been replaced by Entrez Gene.  Entrez Gene is the successor to LocusLink and has two major differences that differentiate it from Locus Link: Entrez Gene is greater in scope (more of the genomes represented by NCBI Reference Sequences or RefSeqs) and Entrez Gene has been integrated for indexing and query in NCBI's Entrez system.

Between-Markers Positional Query

Use this query approach if you are interested in retrieving variations that have been mapped to a specific region of the genome bounded by two STS markers. Other map-based queries are available through the NCBI Map Viewer tool.

ADA Section 508-compliant Link

All links located on the left sidebar of the dbSNP homepage are also provided in text format at the bottom of the page to support browsing by text-based Web browsers. Suggestions for improving database access by disabled persons should be sent to the dbSNP development group [snp-admin@ncbi.nlm.nih.gov].

Flanking Sequence Context DNA or cDNA

Figure 1

Alleles

Table 3. Allele definitions define the class of the (more...)

Table 3. Allele definitions define the class of the variation in dbSNP.

dbSNP variation classa, b	Rules for assigning allele classes	Sample allele definition	Class code in Sybase, ASN.1, and XMLc
Single Nucleotide Polymorphisms (SNPs)a	Strictly defined as single base substitutions involving A, T, C, or G.	A/T	1
Deletion/Insertion Polymorphisms (DIPs)a	Designated using the full sequence of the insertion as one allele, and either a fully defined string for the variant allele or a “-” character to specify the deleted allele. This class will be assigned to a variation if the variation alleles are of different lengths or if one of the alleles is deleted (“-”).	T/-/CCTA/G	2
Heterozygous sequencea	The term heterozygous is used to specify a region detected by certain methods that do not resolve the polymorphism into a specific sequence motif. In these cases, a unique flanking sequence must be provided to define a sequence context for the variation.	(heterozygous)	3
Microsatellite or short tandem repeat (STR)a	Alleles are designated by providing the repeat motif and the copy number for each allele. Expansion of the allele repeat motif designated in dbSNP into full-length sequence will be only an approximation of the true genomic sequence because many microsatellite markers are not fully sequenced and are resolved as size variants only.	(CAC)8/9/10/11	4
Named varianta	Applies to insertion/deletion polymorphisms of longer sequence features, such as retroposon dimorphism for Alu or line elements. These variations frequently include a deletion “-” indicator for the absent allele.	(alu) / -	5
No-variationa	Reports may be submitted for segments of sequence that are assayed and determined to be invariant in the sample.	(NoVariation)	6
Mixedb		Mix of other classes	7
Multi-Nucleotide Polymorphism (MNP)a	Assigned to variations that are multi-base variations of a single, common length.	GGA/AGT	8

Seven of the classes apply to both submissions of variations (submitted SNP assay, or ss#) and the non-redundant refSNP clusters (rs#'s) created in dbSNP.

The “Mixed” class is assigned to refSNP clusters that group submissions from different variation classes.

Class codes have a numeric representation in the database itself and in the export versions of the data (ASN.1 and XML).

Method

Population

Sample Size

There are two sample-size fields in dbSNP. One field is called SNPASSAY SAMPLE SIZE, and it reports the number of chromosomes in the sample used to initially ascertain or discover the variation. The other sample size field is called SNPPOPUSE SAMPLE SIZE, and it reports the number of chromosomes used as the denominator in computing estimates of allele frequencies. These two measures need not be the same.

Population-specific Allele Frequencies

Table 4. Validation status codes summarize the available (more...)

Table 4. Validation status codes summarize the available validation data in assay reports and refSNP clusters.

Validation evidence	Description	Code in database for ss#	Code in FTP dumps for ss#	Code in database for rs#	Code in FTP dumps for rs#
Not validated	For ss#, no batch update or validation data, no frequency data (or frequency is 0 or 1). rs# status code is OR'd from the ss# codes.	0	Not present	0a	Not present
Multiple reporting	Status = 1 for an rs# with at least two ss# numbers; having at least one ss# is validated by a non-computational method. For a ss#, status = 1 if the method is non-computational.	1	1b	1,0b	1
With frequency	Frequency data is present with a value between 0 and 1.	2	2	2	2
Both frequency	For ss#, the method is non-computational and frequency data is present. If the ss# is a single cluster member, then the rs# code is set to 2.	3	3	3/2	3
Submitter validation	Submission of a batch update or validation section that reports a second validation method on the assay.	4	4	4	4

If the rs# has a single ss# with code 1, then rs# is set to code 0.

For a single member rs where the ss# validation status = 1, the rs# validation status is set to 0.

Population-specific Genotype Frequencies

Similar to alleles, genotypes have frequencies in populations that can be submitted to dbSNP.

Population-specific Heterozygosity Estimates

Some methods for detection of variation (e.g., denaturing high-pressure liquid chromatography or DHPLC) can recognize when DNA fragments contain a variation without resolving the precise nature of the sequence change. These data define an empirical measure of heterozygosity when submitted to dbSNP.

Individual Genotypes

dbSNP accepts individual genotypes for samples from publicly available repositories such as CEPH or Coriell. Genotypes reported in dbSNP contain links to population and method descriptions. General genotype data provide the foundation for individual haplotype definitions and are useful for selecting positive and negative control reagents in new experiments.

Validation Information

Submitted SNPs and Reference SNP Clusters

Once a new SNP is submitted to dbSNP, it is assigned a unique submitted SNP ID number (ss#). Once the ss number is assigned, we align the flanking sequence of each submitted SNP to its appropriate genomic contig. If several ss numbers map to the same position on the contig, we cluster them together, call the cluster a “reference SNP cluster”, or “refSNP”, and provide the cluster with a unique RefSNP ID number (rs#). If only one ss number maps to a specific position, then that ss is assigned an rs number and is the only member of its RefSNP cluster until another submitted SNP is found that maps to the same position.

Figure 4. rs7412 has an average heterozygosity of (more...)

Figure 4. rs7412 has an average heterozygosity of 18.3% based on the frequency data provided by seven submissions, and the cluster as a whole is validated because at least one of the underlying submissions has been experimentally validated. rs7412 is annotated as a variation feature on RefSeq contigs, mRNAs, and proteins. Pointers in the refSNP summary record direct the user to additional information on six additional Web sites, through linkout URLs. These Web sites may contain additional data that were used in the initial variation call, or it may be additional phenotype or molecular data that indicate the function of the variation.

Figure 4

New Submissions and the Start of a New Build

Each build starts with a “close of data” that defines the set of new submissions that will be mapped to genome sequence by multiple cycles of BLAST and MegaBLAST for subsequent annotation and grouping of the SNPs into refSNPs. The set of new data entering each build typically includes all submissions received since the close of data in the previous build.

Mapping to a Genome Sequence

When a new genome build is ready, dbSNP obtains the FASTA files for submitted SNPs that were submitted prior to the “close of data”, as well as the FASTA files for the refSNPs in the current build, and then maps the submitted SNPs and refSNPs to the genome sequence using the procedure described in Appendix 2. The refSNP set is also mapped to the previous genome assembly to support users who require older mapping data (e.g. during the production cycle for dbSNP human build 126, The refSNP set was mapped to both human build 36.1 and human build 35.1).

It should also be mentioned that during a build cycle, some organisms have refSNPs mapped to multiple assemblies (e.g. human has two major assemblies: the NCBI Reference Genome assembly and the Celera assembly).

Functional Analysis

Variation Functional Class

We compute a functional context for sequence variations by inspecting the flanking sequence for gene features during the contig annotation process, and do the same for RefSeq/GenBank mRNAs.

Table 5. Function codes for refSNPs in gene features. (more...)

Table 5. Function codes for refSNPs in gene features. ^a

Functional class	Description	Database code
Locus region	Variation is within 2 Kb 5′ or 500 bp 3′ of a gene feature (on either strand), but the variation is not in the transcript for the gene. This class is indicated with an “L” in graphical summaries. As of build 127, function code 1 has been modified into a two digit code that will more precisely indicate the location of a SNP. The two digit code has function code 1as the first digit, which will keep the meaning as described above, and 3 or 5 as the second digit, which will indicate whether the SNP is 3’ or 5’ to the region of interest. See function codes 13 and 15 in this table.	1
Coding	Variation is in the coding region of the gene. This class is assigned if the allele-specific class is unknown. This class is indicated with a C in graphical summaries. This code was retired as of dbSNP build 127.	2
Coding-synon	The variation allele is synonymous with the contig codon in a gene. An allele receives this class when substitution and translation of the allele into the codon makes no change to the amino acid specified by the reference sequence. A variation is a synonymous substitution if all alleles are classified as contig reference or coding-synon. This class is indicated with a C in graphical summaries.	3
Coding-nonsynon	The variation allele is nonsynonymous for the contig codon in a gene. An allele receives this class when substitution and translation of the allele into the codon changes the amino acid specified by the reference sequence. A variation is a nonsynonymous substitution if any alleles are classified as coding-nonsynon. This class is indicated with a C or N in graphical summaries. As of build 128, function code 4 has been modified into a two digit code that will more precisely indicate the nonsynonymous nature of the SNP. The two digit code has function code 4 as the first digit, which will keep its original meaning, and 1, 2, or 4 as the second digit, which will indicate whether the SNP is nonsense, missense, or frameshift. See function codes 41, 42, and 44 in this table	4
mRNA-UTR	The variation is in the transcript of a gene but not in the coding region of the transcript. This class is indicated by a “T” in graphical summaries. As of build 127, function code 5 has been modified into a two digit code that will more precisely indicate the location of a SNP. The two digit code has function code 5 as the first digit, which will keep its original meaning, and 3 or 5 as the second digit, which will indicate whether the SNP is 3’ or 5’ to the region of interest. See function codes 53 and 55 in this table.	5
Intron	The variation is in the intron of a gene but not in the first two or last two bases of the intron. This class is indicated by an L in graphical summaries.	6
Splice-site	The variation is in the first two or last two bases of the intron. This class is indicated by a “T” in graphical summaries. As of build 127, function code 7 has been modified into a two digit code that will more precisely indicate the location of a SNP. The two digit code has function codes 7 as the first digit, which will keep its original meaning, and 3 or 5 as the second digit, which will indicate whether the SNP is 3’ or 5’ to the region of interest. See function codes 73 and 75 in this table.	7
Contig-reference	The variation allele is identical to the contig nucleotide. Typically, one allele of a variation is the same as the reference genome. The letter used to indicate the variation is a C or N, depending on the state of the alternative allele for the variation.	8
Coding-exception	The variation is in the coding region of a gene, but the precise location cannot be resolved because of an error in the alignment of the exon. The class is indicated by a C in graphical summaries.	9
NearGene–3	Function Code 13, where: 1=locus region (see function code 1 in this table) 3=SNP is 3’to and 0.5kb away from gene	13
NearGene–5	Function Code 15, where: 1=locus region (see function code 1 in this table) 5= SNP is 5’ to 2kb away from gene	15
Coding-nonsynonymous nonsense	Function Code 41, where:4 =Coding- nonsynonymous (see function code 4 in this table) 1 = Nonsense (changes to the Stop codon)	41
Coding-nonsynonymous missense	Function Code 42, where:4 =Coding- nonsynonymous (see function code 4 in this table) 2 = missense (alters codon to make an altered amino acid in protein product)	42
Coding-nonsynonymous frameshift	Function Code 44, where:4 =Coding- nonsynonymous (see function code 4 in this table) 4 = frameshift (alters codon to make an altered amino acid in protein product)	44
UTR–3	Function code 53, where: 5= UTR (untranslated region: see function code 5 in this table)3= SNP located in the 3’ untranslated region	53
UTR–5	Function code 55, where: 5= UTR (untranslated region: see function code 5 in this table)5(as the second digit)= SNP located in the 5’ untranslated region	55
Splice–3	Function code 73, where: 7=splice site (see function code 7 in this table)3=3’ acceptor dinucleotide	73
Splice–5	Function code 75, where: 7=splice site (see function code 7 in this table)5=5’ donor dinucleotide	75

Most gene features are defined by the location of the variation with respect to transcript exon boundaries. Variations in coding regions, however, have a functional class assigned to each allele for the variation because these classes depend on allele sequence.

Table 5 defines variation functional classes. We base class on the relationship between a variation and any local gene features. If a variation is near a transcript or in a transcript interval, but not in the coding region, then we define the functional class by the position of the variation relative to the structure of the aligned transcript. In other words, a variation may be near a gene (locus region), in a UTR (mRNA-utr), in an intron (intron), or in a splice site (splice site). If the variation is in a coding region, then the functional class of the variation depends on how each allele may affect the translated peptide sequence.

Typically, one allele of a variation will be the same as the contig (contig reference), and the other allele will be either a synonymous change or a non-synonymous change. In some cases, one allele will be a synonymous change, and the other allele will be a non-synonymous change. If either allele in the variation is a non-synonymous change, then the variation is classified as non-synonymous; otherwise, the variation is classified as a synonymous variation. The primary functional classifications are as follows:

• The functional class is noted as Contig Reference when the allele is identical to the contig (contig reference), and hence causes no change to the translated sequence.

• The functional class is noted as synonymous substitution when an allele that is substituted for the reference sequence yields a new codon that encodes the same amino acid.

• The functional class is noted as non-synonymous substitution when an allele that is substituted for the reference sequence yields a new codon that encodes a different amino acid.

• The functional class is noted as coding when a problem with the annotated coding region feature prohibits conceptual translation. The coding notation is based solely on position in this case.

Because functional classification is defined by positional and sequence parameters, two facts emerge: (a) if a gene has multiple transcripts because of alternative splicing, then a variation may have several different functional relationships to the gene; and (b) if multiple genes are densely packed in a contig region, then a variation at a single location in the genome may have multiple, potentially different, relationships to its local gene neighbors.

Population Diversity Data

The best single measure of a variation's diversity in different populations is its average heterozygosity. This measure serves as the general probability that both alleles are in a diploid individual or in a sample of two chromosomes. Estimates of average heterozygosity have an accompanying standard error based on the sample sizes of the underlying data, which reflects the overall uncertainty of the estimate. dbSNP’s computation of average heterozygosity and standard error for RefSNP clusters is available online. Please note that dbSNP computes heterozygosity based on the submitted allele frequency for each SNP. If the frequency data for a SNP is not submitted, we cannot compute the heterozygosity value, and therefore the refSNP report will show no heterozygosity estimate.

Additional population diversity data include population counts, individuals sampled for a variation, genotype frequencies, and Hardy Weinberg probabilities.

Build Resource Integration

We annotate the non-redundant set of variations (refSNP cluster set) on reference genome sequence contigs, chromosomes, mRNAs, and proteins as part of the NCBI RefSeq project (Chapter 18). We compute summary properties for each refSNP cluster, which we then use to build fresh indexes for dbSNP in the Entrez databases, and to update the variation map in the NCBI Map Viewer. Finally, we update links between dbSNP and dbMHC, OMIM, Homologene, the NCBI Probe database, PubMed, UniGene, and UniSTS.

Public Release

Public release of a new build involves an update to the public database and the production of a new set of files on the dbSNP FTP site. We make an announcement to the dbsnp-announce mailing list when the new build for an organism is publicly available.

Links from SNP Records to Submitter Websites

The SNP database supports and encourages connections between assay records (submitted SNP ID numbers, or ss numbers) and supplementary data on the submitter's Web site. This connection is made using the LINKOUT field in the SNPassay batch header. LinkOut URLs are base URLs to which dbSNP can append the submitter's ID for the variation to construct a complete URL to the specific data for the record. We provide LinkOut pointers in the batch header section of SNP detail reports and in the refSNP report cluster membership section.

Links within NCBI

We make the following connections between refSNP clusters and other NCBI resources during the contig annotation process:

Schema: The dbSNP Physical Model

A schema is a necessary part of constructing your own copy of dbSNP because it is a visual representation of dbSNP that shows the logical relationship between data in dbSNP. It is available as a printable PDF file from the dbSNP FTP site.

Data in dbSNP are organized into “subject areas” depending on the nature of the data. The data dictionary currently includes a description of all the tables in dbSNP as well as tables of columns and their properties. Foreign keys are not enforced in the physical model because they make it harder to load table data asynchronously. In the future, we will add descriptions of individual columns. The data dictionary is also available online from the dbSNP Web site.

Stepwise Procedure for Creating a Local Copy of dbSNP

• 1

  Prepare the local area.

(check available space, etc.)

• 2

  Download the schema files.

• a

  Download the following files from the dbSNP shared_schema sub-directory: dbSNP_main_table, dbSNP_main_index_constraint, and all the files in the shared_data sub-directory. Together, the files from both of these sub-directories will allow you to create tables and indices for the dbSNP_main_table.

• b

  Go to the organism_schema subdirectory, and select the organism for which you wish to create a database. For the purpose of this example, human_9606 has been selected. Once human_9606 is selected, you will be directed to the human organism_schema sub-directory. Download all of the files contained in this subdirectory.

• c

  Go to the organism_data subdirectory, and select the organism for which you wish to create a database. For the purpose of this example, human_9606 has been selected. Once you select human_9606, you will be directed to the human organism_data sub-directory. Download all of the files contained in this subdirectory.

A user must always down load the files located in the most recent versions of the shared_schema and shared_data sub-directories in addition to any organism specific content.

Save all the files in your local directory and decompress them.

• 3

  Create the dbSNP_main_table

• a

  From the shared_schema sub-directory, use the dbSNP_main_table file to create tables, and use the dbSNP_main_index_constraint files to create indices for the dbSNP main database.

• b

  Load all of the bcp files located in the shared_data sub-directory into the dbSNP_main_table you just created using the data-loading tool of your database server (e.g., bcp for Sybase). See the sample FTP protocol and sample Unix C Shell script (below) for directions.

• c

  Create indices by opening the dbSNP_main_index_constraint.sql file. If you are using a database server that provides the isql utility, then use the following command:

isql -S <servername> -U usename -P password -i dbSNP_main_index_constraint.sql

• 4

  Create the organism specific database

Once the dbSNP_main_table has been created, create the organism specific database using the files in your specific organism’s organism_schema and organism_data subdirectories. Human_9606 will be used for the purpose of this example:

• a

  Create the human_9606 database using the following files found in the human_9606 organism_schema: human_9606_table.sql.gz, human_9606_view.sql.gz, human_9606_index_constraint.sql.gz,

and human_9606_foreign_key.sql.gz

• b

  Load all of the bcp files located in the shared_data sub-directory into the human_9606 database you just created using the data-loading tool of your database server (e.g., bcp for Sybase). See the sample FTP protocol and sample Unix C shell script (below) for directions.

• 5

  Sample FTP Loading protocol.

• a

  Type ftp -i ftp.ncbi.nih.gov (Use "anonymous" as user name and your email as your password).

• b

  Type: cd snp/database

• c

  To get dbSNP_main for shared tables and shared data: Type ls to see if you are in the directory with the right files. Then type “cd shared_schema” to get schema file for dbSNP_main, and finally, type “cd shared_data” to get the data for dbSNP_main.

• d

  Type binary (to set binary transfer mode).

• e

  Type mget *.gz (to initiate transfer). Depending on the speed of the connection, this may take hours since the total transfer size is gigabytes in size and growing.

• f

  To decompress the *.gz files, type gunzip *.gz. (Currently, the total size of the uncompressed bcp files is over 10 GB).

• 6

  Use scripts to automate data loading.

• a

  Located in the loadscript subdirectory of the dbSNP FTP site, there is a file called cmd.create_local_dbSNP.txt that provides a sample UNIX C shell script for creating a local copy of dbSNP_main and a local copy of a specific organism database using files in the shared_schema, and the organism_schema sub-directories.

• b

  Also in the the loadscript subdirectory of the dbSNP FTP site, there is a file called cmd.bulkinsert.txt that provides a sample UNIX C shell script for loading tables with files located in shared_data and organism_data sub-directories.

• 7

  Data integrity (creating a partial local copy of dbSNP).

dbSNP is a relational database. Each table has either a unique index or a primary key. Foreign keys are not reinforced. There are advantages and a disadvantage to this approach. The advantages are that this approach makes it easy to drop and re-create the table using the dbSNP_main_table, which then makes it possible to create a partial local copy of dbSNP. For example, if you are interested only in the original submitted SNP and their population frequencies, and not in their map locations on NCBI genome contigs or GenBank Accession numbers (both are huge tables), then these tables can be skipped (i.e., SNPContigLoc and MapLink). Please remember that mapping tables such as SNPContigLoc will have a build ID prefix and suffix included in its file name. (e.g. SNPContigLoc will be b125_SNPContigLoc_35_1for SNP build 125, and NCBI contig build 35 version 1). Of course, to select tables for a particular query, the contents of each table and the dbSNP entity relationship (ER) diagram need to be understood. The disadvantage of un-reinforced references is that either the stored procedures or the external code needs to be written to ensure the referential integrity.

ASN.1

The docsum_2005.asn file is the ASN structure definition file for ASN.1 and is located in the /specs subdirectory of the dbSNP FTP site. The 00readme file, located in the main dbSNP FTP directory, provides information about ASN.1 data structure and data exchange. ASN.1 text or binary output can be converted into one or more of the following formats: flatfile, FASTA, docsum, chromosome report, RS/SS, and XML. To convert from ASN.1 to another format, request ASN.1 output from either the dbSNP FTP site or the dbSNP batch query pages, and use dstool (located in the “bin” directory of the dSNP FTP site) to locally convert the output into as many alternative formats as needed.

XML

The XML format provides query-specific information about refSNP clusters,

as well as cluster members in the NCBI SNP Exchange (NSE) format. The XML schema is located in the docsum_2005.xsd file, which is housed in the /specs sub-directory of the dbSNP FTP site. A human-readable text form of the NSE definitions can be found in docsum_2005.asn, also located in the /specs sub-directory of the dbSNP FTP site.

FASTA: ss and rs

The FASTA report format provides the flanking sequence for each report of variation in dbSNP, as well as all the submitted sequences that have no variation. ss FASTA contains all submitted SNP sequences in FASTA format, whereas rs FASTA contains all the reference SNP sequences in FASTA format. The FASTA data format is typically used for sequence comparisons using BLAST. BLAST SNP is useful for conducting a few sequence comparisons in the FASTA format, whereas multiple FASTA sequence comparisons will require the construction of a local BLAST database of FASTA formatted data and the installation of a local stand-alone version of BLAST.

rs docsum Flatfile

The rs docsum flatfile report is generated from the ASN.1 datafiles and is provided in the files "/ASN1_flat/ds_flat_chXX.flat". Files are generated per chromosome (chXX in file name),as with all of the large report dumps. Because flatfile reports are compact, they will not provide you with as much information as the ASN.1 binary report, but they are useful for scanning human SNP data manually because they provide detailed information at a glance. A full description of the information provided in the rs docsum flatfile format is available in the 00readme file, located in the SNP directory of the SNP FTP site.

Chromosome Reports

The chromosome reports format provides an ordered list of RefSNPs in approximate chromosome coordinates. Chromosome reports is a small file to download but contains a great deal of information that might be helpful in identifying SNPs useful as markers on maps or contigs because the coordinate system used in this format is the same as that used for the NCBI genome Map Viewer. It should also be mentioned that the chromosome reports directory might contain the multi/ file and/or the noton/ files. These files are lists (in chromosome report format) of SNPs that hit multiple chromosomes in the genome and those that did not hit any chromosomes in the genome, respectively. A full description of the information provided in the chromosome reports format is available in the 00readme file, located in the SNP directory of the SNP FTP site.

Genotype Report

The dbSNP Genotype report shows strain-specific genotype information for model organisms, and contains a genotype detail link as well as a genotype XML link. The genotype detail link will provide the user with submitter and genotype information for each of the submitted SNPs in a refSNP cluster of interest, and the genotype XML link will allow the user to download the reported data in the Genotype Exchange XML format, which can be read by either Internet Explorer or Netscape browsers. XML dumps via the dbSNP ftp server provide the same content for all genotype data in dbSNP by organism and chromosome.

Organism-specific Genome Mapping

The refSNP(rs) and submitted SNP (ss) mapping process is a multi-step, computer-based procedure that begins when refSNP and submitted SNP FASTA sets are aligned to the most recent genome assembly using BLAST or MegaBLAST. Repeat masking during this process is accomplished automatically using the BLAST/MegaBLAST “Dust” option. To increase alignment stringency, multiple cycles of BLAST/MegaBLAST are employed, where the word size limit is reduced from 64 in the first cycle to16 by the final cycle. MegaBLAST parameters are set to a default position with the exception of a seeding hit suppression parameter(e.g. Parameter="-U F -F 'mL' -J T -X 10 -r 1 -q -3 -W 64 -m 11 -e 0.01")that suppresses seeding hits but allows extension through regions of lowercase sequence. The quality of each alignment is determined using an Alignment Profiling Function developed specifically for this purpose. Alignments are selected based on a variable stringency threshold that ranges from 70% alignment to 50% alignment. If the alignment profiling function indicates the quality of the alignment is below a 70% alignment threshold, the alignment is discarded, although an alignment threshold of 50% is sometimes used in case there are gaps in the sequence.

The BLAST/MegaBLAST output of ASN.1 binary files of local alignments is then analyzed by an algorithm (“Globalizer”) that sorts those local alignments that do not fit the dbSNP alignment profile criteria (defined by position and proximity to one another) to create a “Global Alignment” — a group of local alignments that lay close to one another on a sequence. If the global alignment is greater than or equal to a pre-determined percentage of the flanking sequence, it is accepted as a true alignment between the refSNP or submitted SNP and the genome assembly. The Globalizer is especially helpful when refSNPs or submitted SNPs based on mRNA sequence are being aligned to the genome assembly. In such a case, the MegaBLAST/BLAST ASN.1 binary output contains many small alignments that will not map to the genome assembly unless they undergo the “globalization” process.

The ASN.1 binary output of “Globalizer” is then processed by a program called “Hit Analyzer”. This program defines the alleles and LOC types for each hit, and also determines the map position by using the closest map positions on either side of the SNP to establish the hit location. The text output of “Hit Analyzer” is then processed by the “Hit Filter”, which filters out paralogous hits and uses multiple strategies to select only those SNPs that have the greatest degree of alignment to a particular contig. The output from the “Hit Filter” is then placed into a map.bcp file and is processed by the “SnpMapInfo” program, which creates an MD5 signature for each SNP that is representative of all the positional information available for that SNP. The MD5 signature is then placed in the SNP MAP INFO file, which is then loaded into dbSNP.

RefSNPs and submitted SNPs are analyzed against GenBank mRNA, RefSeq mRNA, and GenBank clone accessions using a similar procedure to that described in the above paragraphs.

Once the all the results from previous steps are loaded into dbSNP, we perform cluster analysis using a program called “SNPHitCluster” which analyses SNPs having the same signatures to find candidates for clustering. If an MD5 signature for a particular SNP is different from the MD5 signature of another SNP, then the hits for those two SNPs are different, and therefore, the SNPs are unique and need not be clustered. If an MD5 signature of a particular SNP is the same as that of another SNP, the two SNPs may have the same hit pattern, and if after further analysis, the hit patterns are shown to be the same, the two SNPs will be clustered.