AceView downloads

The .gff file provides the structure of the genes at the exclusion of the 'cloud', i.e. the coordinates of the genes elements. The tar.gz contains one file per chromosome, MT is mitochondria.

The gff format is nicely defined by Jim Kent on the UCSC genome browser.

Aim of this file:

To allow genome-wide resources depicting genes to represent nicely the properties of the alternatively spliced variants seen by AceView upon mapping and clustering of the millions public mRNA sequences on the genome. This table provides, for each mapped gene, the GeneID, the coordinates of the gene on the genome, and a description of all its cDNA-supported introns. It then provides, for each mRNA variant, the coordinates in the gene, the intron signature and information on completeness at both the 5’ and 3’ ends: putative promoters and confirmed alternative polyadenylation sites are annotated. For putative protein-coding mRNAs, information on the protein, its location in the transcript, its completeness, which start codon is possibly used and which cDNA clone(s)/accessions encode a protein identical to the predicted protein is given.

Format:

To avoid redundancy and to keep the table simple, we let the content of each column depend on the TYPE of data, given in the first column of each line. There are currently 5 types, that do appear in successive order in the table: GENE, INTRON (in the above gene), mRNA (in the gene above; there are as many mRNA lines under a gene as there are alternative variants, INTRONS (in the mRNA above it), PROTEIN (in the mRNA above, if applicable).
This format is versatile and will allow any desirable extensions.
The table is tab delimited; multi-valued columns are semi-column delimited (';').

Example :
GENE AMELY 266 Y 6785429 6777320 8110
INTRON 1 gt_ag 57;1407 5 BC069138;M86933;NM_001143;BC074976;BC074977
INTRON 2 gt_ag 1474;3974 7 AY487421;BC069138;M86933;NM_001143;BC074976;BC074977;BG198114
INTRON 3 gt_ag 4023;5118 7 AY487421;BC069138;M86933;NM_001143;BC074976;BC074977;BG198114
INTRON 4 gt_ag 4023;5251 7 AY487421;BC069138;M86933;NM_001143;BC074976;BC074977;BG198114
INTRON 5 gt_ag 5161;5251 1 AY487421
INTRON 6 gt_ag 5297;5565 7 AY487421;BC069138;M86933;NM_001143;BC074976;BC074977;BG198114
INTRON 7 gt_ag 5992;7949 7 AY487421;BC069138;M86933;NM_001143;BC074976;BC074977;BG198114
mRNA AMELY.aAug05 266 1420 7955 - - AY487421
INTRONS 5 2;3;5;6;7 1474;3974;7 4023;5118;7 5161;5251;1 5297;5565;7 5992;7949;7
PROTEIN 4 1420 7955 NH2_partial Stop
mRNA AMELY.bAug05 266 1 8110 1 7983;8110 M86933
INTRONS 5 1;2;4;6;7 57;1407;5 1474;3974;7 4023;5251;7 5297;5565;7 5992;7949;7
PROTEIN 4 1420 7955 ATG Stop BC069138;M86933;NM_001143;BC074976;BC074977

Content of the columns, per TYPE:

GENE

• Column 1: the name of the gene in AceView.
  If there is an official name, the AceView name is the official gene name, except in the infrequent cases where there is evidence that at least one transcript shares introns with transcripts from two adjacent Entrez genes with different official names, say GENEa and GENEb. In this case, the gene is more complex than previously thought, and we name it GENEaANDGENEb, concatenating the official gene names. Each transcript then knows if its best contact is to GENEa or to GENEb, and bears that GeneID. If there is no official gene name, we use a Pfam-derived name.dot.number, else a purely computer generated AceView name. We try to track all gene names from release to release.
• Column 2: the Entrez GeneID, uniquely inherited in this order by contact with an NCBI gene model aligned by NCBI on the reference genome, otherwise through a RefSeq model aligned in the AceView gene, otherwise through one of the 'reference mRNAs' in Entrez gene. A number of AceView genes do not have a corresponding Entrez GeneID yet, and, as explained above, some AceView genes merge multiple genes with different geneIDs, hence this column may be Null (-) or multi-valued.
• Column 3: The chromosome number (1 to 22, X and Y; NT_* denotes unmapped contigs and follows the NCBI nomenclature)
• Column 4: The coordinate of the 5’ end of the gene on the chromosome, in bp, as currently supported by cDNAs in the public domain (the gene may be partial, and nearby genes on the same strand may become merged by new cDNA sequences in a future release.)
• Column 5: The coordinate of the 3’ end of the gene on the chromosome, in bp. Note that comparing the values in column 4 and 5 gives the strand of the gene: if the value in column 5 is greater than the value in column 4, the gene is encoded on the forward strand; otherwise it is on the reverse strand.
• Column 6: the extent of the gene on the genome, in bp. Cumulatively, that is the extent of the genome transcribed in pre-messengers that will mature into mRNAs from this gene.

INTRON:
« Introns » are here taken in a broad sense to denote any discontinuity in the cDNAs alignments. The coordinates of exons can be deduced from the table and are not given explicitly. Because the ends of the terminal 5' and 3' exons are biologically variable and often partial, they are called alternative exons and appear artificially more numerous than alternative introns. More than 99% of the objects under INTRON are introns defined by perfect alignments of cDNAs over at least 8 bases on each side of the intron, 98% of all are standard introns with [gt-ag] or [gc-ag] boundaries. 1.7% are non-standard or fuzzy (imperfectly defined), the remaining 0.3% correspond to alignment gaps.

Following our analysis in GOLD, we now annotate all human introns below 60 bp as structural size polymorphisms and note they most often occur in the 3’UTRs. The relatively low level of non-standard introns we report reflects the fact that we recognize and flag cDNA clone anomalies such as partial insert deletions or rearrangements carefully (in about 5% of the cDNAs). To represent a variant with a non-standard intron, we demand either that three or more cDNAs support the non-standard intron, or that the cDNA harboring this non-standard intron also brings a unique standard alternative intron or exon.

AceView currently annotates close to twice the number of standard introns in RefSeq or EBI, yet all AceView introns in the table are supported by cDNA sequence alignments that are locally perfect (over 16 bp), hence unambiguous. The number of introns continues to grow almost linearly with the number of cDNA sequences deposited in the public databases, a sure sign that we are still far from saturation and that the transcriptome is truly more complex than what we describe today, even in AceView!
- Column 1: ordinal number of the intron
- Column 2: intron boundaries, either standard gt-ag or gc-ag, or any other sequence, fuzzy or gap
- Column 3: coordinates of the first and last base of the intron, relative to the gene (base 1 is the 5’ end of the gene).
- nombre de séquences soutenant exactement cet intron avec 8 bases exactes de chaque coté;
- liste des accessions du gène soutenant exactement l’intron
- First the number of introns (standard, with gt-ag or gc-ag intron boundaries); then the number of other non-
standard introns; then the number of alignment or sequencing gaps if any. The three numbers are semi-column delimited.

The intron-exon table provides the elements of the multiexon transcripts (those for which the first number in column 11 above is not 0), also in the coordinates of the gene.
• Columns 1: ordinal number of the intron in the transcript
• Column 2: ordinal number of the intron in the gene? To help the drawing at GeneCards. Then maybe simply give the intron signature (concatenated intron numbers)
• Columns 3 and 4: coordinates of the first and last base of this intron in the gene
• Column 5: intron boundaries. The first two and last two bases of the intron, almost always gt-ag, some gc-ag, and a few at-ac thought to be U12-dependent.
• Column 6: number of clones supporting each intron in the entire gene
• Column 7: example clones allocated to this variant (semi-column delimited)

The position of each AceView transcript, and whether or not each variant is complete on the 5’ side (putative promoter) and 3’ side. On the 3’ side, we now annotate carefully the multiple alternative polyadenylation sites.
A second table details the “intron exon structure” of the spliced variants and provides the degree of support (number of clones, total or specific) for each intron. their confirmed/validated/well supported/secure starts and ends
- All other coordinates in this table are given relative to the gene
- Column 6: the variant name. Note: the variant encoding the largest predicted protein is called a, then b and so on. Alternative variants may change from release to release as new cDNAs are submitted, but the variant name includes the date of the release aDate, bDate…, which makes it unambiguously attached to a single sequence.
- Column 7 and 8: The coordinates in the gene of the first and last base of the transcript (base 1 is the first base of the gene, i.e. the first base of the 5’-most transcript of the gene)
- Column 9: A well supported 5’ end, indicating a promoter. When there is evidence that the transcript is likely
complete on the 5’ side, the coordinate in the gene of the validated transcript start is given. There is at most one 5’end per variant. Elements to decide on 5’ completion include close agglomeration of many 5’ ends of cDNAs, the presence of a clone from a cap-selected or full-length enriched library within the first 50 bases of the transcript (because the actual start of transcription may biologically span up to 61 bp, according to is Suzuki et al, EMBO Rep. 2001 May;2(5):388-93). If no such evidence exists, we also consider as putatively complete on the 5’ side a transcript encoding a protein of good quality, bounded on the NH2 side by an in-frame Stop; in such a case, the real end of the transcript may be somewhat upstream of the current 5’end, but introns in 5’UTR are infrequent. Any transcript for which there is no evidence for completion on the 5’ end has NULL in this column.
- Column 10: the coordinates in the gene of all well supported (alternative) polyadenylation sites. In AceView, in order to reduce the apparent number of transcripts, we merge transcript variants that share the same introns and only differ in the length of the last exon i.e. in the position of their poly adenylation sites. We now indicate where the well supported alternative ends are. Factors used to determine the likely 3’ ends involves clustering the known 3’ ends, either from cDNAs or ESTs where the polyA is visible in the sequence, or from groups of reverse-aligned (usually 3’) ESTs whose 3’ ends cluster in a short area, or from extremities of the mRNAs (more precisely reference mRNA quoted in Entrez Gene) that have been submitted to GenBank with the mention ‘Complete CDS’. The ‘clusters of 3’ ends’ are then analyzed for the presence of A-rich stretches just downstream of each cluster in the genome, suggesting that the clones, usually prepared using oligo dT aiming at the poly A actually hybridized to a region rich in A in the mRNA and/or the genome. When such an A-rich region is found, all clones in the cluster are labeled as ‘internal priming in A rich region’, and the 3’ end is not confirmed. Finally, a search for the AATAAA or any of its single letter variants in the correct position, less than 40bases before the cluster of 3’ ends, allows to define the most probable ‘true’ 3’ ends. Contrary to UniGene, we let singlebase variants of the main AATAAA signal contribute because we found that altogether they account for just about half of the signals in unambiguously complete transcripts.
Some transcripts have multiple alternative polyadenylation sites (multiple confirmed 3’ ends sites, this column
may be multi-valued), while some are incomplete on the 3’ end and have no such site annotated.
- Column 11: the number of introns in the transcript, first the number of standard introns (with gt-ag or gc-ag
intron boundaries); then the number of other non-standard introns; then the number of alignment or sequencing gaps if any. The three numbers are semi-column delimited.
- Column 12: the protein-coding score. Variants with scores greater than 1 are likely protein coding, variants with 0 or negative scores are probably not. The magnitude of the score depends on the size, conservation and
annotation of the protein as well as on the position of the CDS relative to the intron scars. Scores are also used to select the ‘best predicted protein’, whose annotation is displayed on the web (more details here).
- Column 13 and 14: the coordinates of the ‘best predicted protein’ in the gene. Some proteins are partial. Many are complete and start at ATG/Met. Some of the complete proteins gain at least 60 aminoacids relative to the standard ATG if they start at a single letter variant of ATG, following the Kozak rule: those are hypothesized to possibly encode the longer variant, eventually in addition to a shorter isoform starting on the first Met. Only the longest variant is annotated in AceView, but since most of the annotation is local, it is easy to deduce the annotation of the shorter form by examining the graphical representation of the variant.
- Column 15: the number of standard introns outside the best protein. Unambiguously coding transcripts most often have no intron outside the CDS; some have one in the 5’UTR. It has been showed that some transcripts with introns downstream of the CDS might be targeted for one round of translation, followed by mRNA degradation by the nonsense mediated degradation NMD pathway, but the generality of the phenomenon remains to be demonstrated, and many exceptions have been documented at the functional and molecular level.

The first line of the table reads:
Gene Entrez_GeneID chromosome 5’end_of_gene 3’end_of_gene variant_name First_base_in_gene

Last_base_in_gene defined_5’end(promoter) defined_3’ends(polyA) #introns(standard_other_gap)

protein_coding_score best_protein_begins best_protein_ends #standardIntrons_outsideCDS

Variant_name First intron begins first intron ends intron boundaries totalNumberOfSupportClones

 

example clones allocated to this variant supporting this intron (semi-column delimited? Or one line per accession with repeats of the rest of the line?)
mRNA
les colonnes sont :
- nom du mRNA;
- GeneID (en général single-valued ici);
- coordonnée de la première base dans le gène ;
- coordonnée de la dernière base dans le gène;
- coordonnée du vrai 5’, si le mRNA est complet (indique un promoteur)/ - si le mRNA ne présente pas d’évidence

d’être complet ;
- coordonnées du ou des sites confirmes de polyadenylation, éventuellement multivalue/ - si pas de site de

polyadenylation confirme;
- exemples de clones soutenant le mRNA (choisis comme couvrant le mRNA ou au moins le CDS et ayant la séquence la

plus proche de celle du génome sous-jacent)

INTRONS (dans le mRNA dessus)
les colonnes sont :
- nombre d’introns dans le mRNA de la ligne au dessus;
- signature des introns du mRNA (numéros des introns decrits dans INTRON, concaténés) ;
- chacune des colonnes suivantes contient pour chaque intron successivement un triplet de chiffres: coordonnée de la

première base de l’intron dans le gène, dernière base de l’intron, nombre d’accession soutenant cet intron, les trois

chiffres sont séparés par des point virgules. Ca devrait etre utile pour tes dessins.

PROTEIN (meilleur protéine prédite dans le mRNA ci-dessus, si le mRNA semble codant)
les colonnes sont :
- le grade de la protéine (un système de gradation de qualité des protéines utilise pour choisir la meilleur

protéine codée par le transcrit ; je vais decrire ca en anglais pour la prochaine release. Un mRNA pour lequel la

meilleure protéine a un grade inferieur a 1 est traite comme non codant ou trop incomplet et n’est pas associe a une ligne

PROTEIN) ;
- coordonnée dans le gène du début de la protéine (bp) ;
- coordonnée dans le gène de la fin de la protéine, incluant le Stop ;
- nature du start de la protéine (soit NH2_partial, soit si le produit est complet, on indique le codon initiateur,

en general ATG, parfois un autre codon qui peut servir d’initiateur (selon Kozak et autres exemples) et qui gagne au moins

60 aminoacides par rapport a l’usage du premier ATG) ;
- nature de la fin de la protéine, Stop ou COOH_partial ;
- liste des clones dont la séquence code exactement pour la protéine, sans aucune substitution d’acide amine

 

The file gives the relationship between the mRNA variant in the fasta files on our downloads site, the AceView gene, the Entrez gene ID and the RefSeq accession when applicable.

• Col 1: mRNA variant, as in the fasta mRNA file
• Col 2: gene name (in AceView; it is the official name at the date of the release whenever there is an official name, else a gene name derived from a significant Pfam protein domain homology, else a computer generated name specific to AceView and maintained from release to release, all in lower case. Please if your gene does not have an official gene name, ask for one before you publish on this gene, as AceView gene names have nothing official.
• Col 3: GeneID (if any). Often a gene is annotated in AceView because it is supported by cDNA sequences, but it is not yet an official gene in Entrez gene, hence it has no GeneID. Also, in some cases, AceView sees one gene where Entrez sees two, so in those cases there will be two GeneID associated to that single AceView gene.
• Col 4: NM included in this variant if any, but mind you, we allow extensions of NMs so the correspondence means inclusion, not necessarily identity.

New, February 8, 2009

Introns are discovered by alignment to the reference genome and annotated in AceView. Our criteria for a cDNA to “support a junction” is that the cDNA sequence exactly matches 16 bp centered on the junction (8 bases in each exon bordering the intron).

This file documents the exon-exon junctions (or intron scars) seen in cDNA sequence accessions from any of the NCBI sequence databases (GenBank, dbEST, Trace, SRA, GEO). It provides, in tab delimited format, the gene, GeneID, intron boundaries, coordinates on the genome of the intron, the ~98 bp exonic sequence around the junction (49 bp of exon 1 followed by 49 bp of exon 2, when available) and finally the RefSeq and AceView transcript models each intron belongs to. The supporting accessions from GenBank/dbEST or Trace are given in an additional file.

 

Format and details:

• file AceViewRefSeqExonsJunctions.txt.gz

The title line reads:

1. #Junction 
2. Gene   
3. GeneId 
4. Intron boundary
5. Chromosome
6. First base of intron   
7. Last base of intron    
8. Strand 
9. Junction position
10. Sequence       
11. In transcript

Col 1: Junction name: chromosome__coordinate_coordinate_build#
We propose that the intron name includes the reference genome build for the organism (e.g. 36) in addition to the chromosomal coordinates. This nomenclature for introns can easily be understood.

Col 2: gene name.
Entrez/HGNC gene names available at AceView release date are systematically used in AceView. Genes absent from the Entrez gene annotation are given Pfam names or all lower case names (easily distinguished from all upper case official human gene names), AceView names are stably maintained across builds. Note that a few AceView genes join two official genes because some cDNA clone bridges them and either shares an intron boundary with both genes and/or has substantial sequence overlap with both: when this occurs, the AceView gene name is a concatenation of the official meaningful gene names (not the LOC#) joined with 'and'.

Col 3: Entrez Gene ID, when available.
Many spliced genes, especially non-protein-coding, are not yet in Entrez gene, so they do not have a GeneID at this time, although they have a stable AceView gene name.
Conversely, some genes in AceView contact models of two genes in Entrez/RefSeq, and some junctions cannot be attributed unambiguously to only one ID. In such a case, both GeneIDs are listed for that junction.

Col 4: intron boundaries: generally gt-ag, some gc-ag. Rare at-ac occur in a few U12 dependent cases, and a large variety of other intron boundaries are used in rare instances.

Col 5: chromosome
The file includes the entire genome, except mitochondria (chromosomes 1 to 22, X and Y). The unattached contigs are given in the NCBI official nomenclature: their name starts with the chromosome name. Note that most of the ribosomal genes and other high copy number sequences are absent or under-represented in the reference genome.

Col 6: coordinate on the chromosome of the first base of the intron (i.e. base immediately after the last base of the first exon)
The first base of the chromosome is called 1.

Col 7: coordinate on the chromosome of the last base of the intron (i.e. base immediately before the first base of the second exon)
Note that comparison of the values in col 3 and 4 gives the strand (col 3 < col 6: mRNA is on the top strand, col 3 > col 4: mRNA is on the bottom strand)

Col 8: strand, + or –

Col 9: length of DNA before the exon junction in the sequence given in column 7.
As we aim at providing 98 bp junction sequences, the junction usually lies at 49, except when the first exon is a first 5' exon in a transcript and is shorter than 49 bases.

Col 10: sequence of the exons junction
Usually 98 bases, but shorter when first or second exon is terminal in the transcript, and its length is less than 49 bp.
Note that AceView transcripts are not extended by concatenation: we prefer partial transcripts to complete but not uniquely supported: there are usually many alternative ways to complete any partial transcript by re-using data from the other transcripts in the gene.

Col 11: list of variants in which the junction is seen, semi-column delimited.
The list includes AceView variants from the April 2007 build and RefSeq entries NM/NR/XM/XR from November 22, 2008.  

 

• File AceViewExonsJunctionsSupport.txt gives the actuall clone and accession support

Col 1: exons junction name, as defined above: chromosome__coordinate_coordinate_GenomeBuildNumber
Col 2: supporting cDNA clone
Col 3: corresponding accession(s). Some clones have both a 5' and a 3' read. Usually the junction is seen in only one of the accessions.

 

 

File: ncbi_*.introns2dna2support is an 8 columns file:

1. # Gene name, in AceView: Entrez/HGNC gene names available at AceView release date are systematically used in AceView. Genes absent from the Entrez gene annotation are given Pfam names or all lower case names (easily distinguished from all upper case official human gene names), AceView names are stably maintained across builds. Note that a few AceView genes join two official genes because some cDNA clone bridges them and either shares an intron boundary with both genes and/or has substantial sequence overlap with both: when this occurs, the AceView gene name is a concatenation of the official meaningful gene names (not the LOC#) joined with 'and'.
2. Intron boundaries: generally gt-ag, some gc-ag. A large variety of other intron boundaries are used in rare instances, for instance at-ac is found in a few cases thought to be U12 dependent.
3. Coordinate in the gene of the first base of the intron. First base of the gene is called base 1.
4. Coordinate in the gene of the last base of the intron
5. Number of exactly supporting clones: Introns are discovered by alignment to the reference genome and annotated in AceView. Our criteria for a cDNA to “support a junction” is that the cDNA sequence exactly matches 16 bp centered on the junction (8 bases in each exon bordering the intron).
6. last 40 bp from the 5' exon
7. first 40 bp from the 3' exon
8. Supporting clones in NCBI Genbank, dbEST or TraceDB

Example line:
"AADACL3"     "gt_ag"           230      3358    2          gagagtcctccattgcatcttccagctgctgttgacatgg            gggatgatatttgagaagctcagaatctgttctatgcccc            "IMAGE:4803719\; NM_001103170"

These files ...genes/AceView.ncbi_*.gene2accession2tissue.main_genes.txt.gz and ...genes/AceView.ncbi_*.gene2accession2tissue.cloud_genes.txt.gz give the relation between all aligned ESTs/mRNAs, the genes, the GeneIDs and the alternative variants, with indication of quality of match, tissue of origin (from GenBank/dbEST), type of AceView gene, for all genes (main and cloud separately)

The file describes the association of each mRNA/EST accession satisfactorily aligned in AceView to the corresponding AceView gene and transcript.
It is ordered by map position, and includes all genes, even the "cloud" genes. We indicate the official gene name and the LocusID when available. We show, for each accession, the alternative variant to which the cDNA contributes, but please be aware that we try to minimize the number of variants to which a given cDNA belongs, to avoid combinatorics in number of AceView transcripts: even when a sequence could equally well participate in two or more variants, we try to assign it to only one. Most alternative variants, except those with structural defects that we filter out, are included.
 The file was generated at the request of various users among which the GeneCards team; it has been made available since build 34 from the “downloads” page. For build 35, we have added the type of AceView gene (main, putative or cloud), tissue for each accession, characteristics and quality of the AceView EST/mRNA to genome match.
For build 34, the file has 5,101,247 lines and 5 columns; for build 35, the file has 6,397,928 lines and 12 columns (we added col 4, 7, 8, 9, 10, 11, 12). 
The current content and format is described below:
 column 1: chromosome position (Position)
This is a volatile name changing with each build, not a stable identifier: it starts with the chromosome name, then coordinate on that chromosome, in basepairs (e.g. 1_1287: chromosome 1, base 1287). It is handy because it allows us to export the table in genome order and can be helpful when you parse the data
column 2: AceView gene name (AceViewID).
This should be stable from release to release, although it evolves as new official gene names are generated. See how we generate the names in AceView: basically, we use official names if they exist, else PFAM derived names, else invented AceView names.
column 3:   set of LocusID or GeneID from Entrez Gene (previously LocusLink LocusID)
Examples:

• 1234 LocusID 1234
• NULL: there is no LocusID yet, because this gene has not yet been modeled in LocusLink/Gene at NCBI.
• 1234;9876  this gene has 2 LocusID, 1234 and 9876. This actually occurs for 1373 human genes in build 34 out of 19,413 AceView genes with a LocusID. In build 35, 1498 human genes out of 24520 AceView genes with a GeneID have more than one GeneID (23,356 GeneID have one or more corresponding AceView gene(s)). Some AceView genes with more than one GeneID correspond to a single gene that was split in the RefSeq/LocusLink model, some to genes producing two different types of proteins, but fused in a single gene in AceView because there is at least one cDNA which bridges the two, even if only through the UTRs. AceView would actually consider this a single (possibly complex) gene, and name it by concatenating the LocusLink/official gene names (see this help and example: PEX19andWDR42A, where clone AGENCOURT_14368013 NIH_MGC_181 Homo sapiens cDNA clone IMAGE:30398500, accession CD518985 creates a nice bridge). Finally, a few correspond to repeated genes in a close tandem arrangements that we may have unintentionally merged.

 column 4: AceView gene type (Type)
Since build 35, the AceView genes have been categorized as Main gene, Putative gene or Cloud gene, as defined here.
 column 5: AceView transcript variant name (Variant)
  This is very often identical to the gene name, except when there is evidence for alternative splicing, then this column shows how we sub-cluster the mRNAs. The relation accession/ alternative variant to which the cDNA contributes is indicated in the next column, but please be aware that we try to minimize the number of variants to which a given cDNA belongs, to avoid combinatorics in number of AceView transcripts: even when a sequence could equally well participate in two or more variants, we try to assign it to only one. On the other hand, it may happen that an accession belongs to several different genes or different variants.
 column 6: mRNA or EST GenBank accession aligned in this gene (Accession)
We give the accession without a version, but for each new AceView release, we provide explicitly the date at which the mRNA/EST data were downloaded from GenBank.
 column 7: Number of basepairs of the accession to be aligned (Length)
This number is derived from the length in basepairs, as found in GenBank, minus the polyA and the eventual vector sequence that AceView recognized and clipped. This is the length in basepairs that we think has to be aligned on the genome.
 column 8: Number of unaligned bases, on the 5’ side (Start)
The value in this column is 0 if the alignment starts at the first base that needs to be aligned (base 1 of the GenBank accession if the vector was correctly clipped by the submitter, first base after the vector if some vector sequence was recognized unclipped in the GenBank accession. For example, the value in this column is 20 if we failed to align the first 20 bases, with no justification: the cDNA might have some structural anomaly or the reference genome might be missing a piece (or rarely, our alignment procedure might be bugged). 
column 9: Number of bases of the mRNA accession actually aligned on the genome in AceView (Ali)
column 10: number of base differences in the aligned region between the cDNA and the genome (Err). A single base transition, transversion, addition or deletion counts as 1.
column 11: Quality of the AceView alignment (Quality), scored from 1 (very best alignments) to 9. The quality factor reflects both the % length aligned and the % differences from the genome.
Scores also depend on whether the sequence is supposed to be high quality (mRNA) or single pass (EST). For mRNAs, the score is measured over the entire length to be aligned (column 7); for ESTs, it is scored on a maximum of 600 bp. Scores follow the empirical chart below: 

%length aligned %bp differences	> 98%		90-98%		80-90%		50-80% or 800 bp		Less than 50% or 800 bp
	mRNA	EST	mRNA	EST	mRNA	EST	mRNA	EST	mRNA	EST
< 0.1%	1	1	2	2	3	3	4	5	7	9
0.1 to 1%	2	2	3	3	4	4	5	6	8	(10)
1 to 2%	3	3	4	4	5	5	6	7	(10)	(11)
2 to 3%	4	4	5	5	6	6	7	9	(11)	(11)
3 to 4%	5	5	6	6	7	7	9	(10)	(11)	(11)
4 to 5%	7	6	8	7	9	8	(10)	(11)	(11)	(11)
> 5%	9	7	(10)	8	(10)	9	(11)	(11)	(11)	(11)

Note: Qualities are initially scored from 1 to 11, but if no structural rearrangement is involved, we ultimately only retain qualities from 1 to 9. This scoring system has been tuned first on the nematode then on unfinished human genome in an empirical way, but it is useful, since qualities are used internally in AceView as a critical step to keep only the best matches during the initial clean-up phase. These numbers may also be used as a quick indicator in the case of other projects, such as GeneCards/GeneTide, to explain some discrepancies between various clustering methods.
 To give ideas on the statistics, for build 35, there are

Aligned at quality	1	2	3	4	5	6	7	8	9	(10)	(11)
#accessions	1,254,893	2,502,199	1,014,606	573,650	387,665	254,449	212,953	114,645	82,866	29,844	559

 column 12: Tissue, as copied from the GenBank submission (not yet systematic). We would prefer to use a standardized vocabulary, such as TissueInfo developed by Lucy Skrabanek and Fabien Campagne, but this is not yet done.

The AceView mapping should improve significantly the information you get from your arrays. We provide mapping of probes from whole genome expression arrays or other microarrays on the AceView genes. Ask us if your favorite arrays are missing.

For human, we routinely remap the Affymetrix HG-U133 Plus 2.0 GeneChip, Agilent Whole Human Genome Oligo Microarray, G4112A and Illumina Human-6 BeadChip, 48K v1.0 platforms.

For mouse, we currently remap Affymetrix 430-2 (44.7 MB) and Illumina MouseWG-6_V1_1_R3 (5.3 MB), WG-6 V2 (6 MB) or MEEBO

Mapping array probes to genes in a systematic fashion is essential for comparing expression results obtained on different platforms, as well as to interpret results from a single platform (for example see the MAQC project). We mapped probes to AceView transcripts and the genome for some of the most used commercial microarray platforms.: in the first release (March 6, 2008), we provide alignments for the Illumina Human-6 BeadChip (48k v1.0), Agilent WHG (G4112A0) and Affymetrix HG-U133 Plus 2.0 GeneChip expression arrays, used in MAQC.
The mapping is performed on the current human genome (build 36) as well as on three annotated transcripts sets, two from NCBI and one from EBI. We used all AceView transcripts (current release April 2007), all RefSeq (NM and XM, but not including the few non coding NR/XR) (downloaded January 2008), or all Ensembl models (downloaded January 2008).
We initially allowed up to 2 mismatches (base difference or indel) for Agilent 60-mer probes and Illumina 50-mer, and up to 1 mismatch for Affymetrix 25-mer probes.
 You may ask for more data of that kind as we are willing to similarly map any platform which provides the sequence of their probes to the public. 

The format is defined below, all files are ordered on the name of the probe (column 4):
Column 1: name of target mRNA (from AceView, RefSeq or Ensembl), or of chromosome if mapping is to reference genome
Column 2: (t1) coordinate of the first base of the match on the target mRNA
Column 3: (t2) coordinate of the last base of the match on the target. (t1 and t2 also give the strand)
Column 4: name of the arrayed probe
Column 5: (p1) coordinate of the first base of the match on the arrayed probe
Column 6: (p2) coordinate of the last base of the match on the probe
Column 7: number of uncalled bases (N) in probe [rare; in some control probes]
Column 8: number of mismatches between probe and target. One mismatch is a difference, an insertion or deletion affecting a single nucleotide
Column 9: coordinate in the probe of the start of the longest exact match
Column 10: coordinate of the end of the longest exact match
Column 11: length of the longest exact match
Column 12: position of the first mismatch
Column 13: type of the first mismatch: may be base_in_transcript > base_in_probe (e.g. a>g) OR +base for extra base in probe (insertion) OR -base for base missing in probe (deletion)
Column 14: only in Affymetrix, indicates the probeset

-----------

from March 2009:

We are happy to remap probes from any (expression) array of interest to the three main transcriptome annotations, NCBI RefSeq, Ensembl, AceView, and to the reference genome.  Depending on the protocol used to prepare the cDNA sample, one may keep the strand information or lose it (for instance because there was a PCR step involved), so we provide the mapping for both situations, as we indicate hits on the mRNA sense or antisense strand. In our mapping, we allow for 10% mismatches (indels or single base variations), which may seem high. Yet the number of mismatches for each probe mapped is indicated in the table, so that users can use the threshold they view as more fit in their analyses.

• For RefSeq, we provide two flavors of mapping, either limited to the more curated NM/NR set or to all RefSeq, including XM/XR models.
• For EBI Ensembl, we use all models, including the ab initio models.
• For AceView, which summarizes the alignable cDNA sequences in the public repositories, we use our latest public version. Unlike Ensembl, AceView does not use any ab initio or predictive elements. Unlike RefSeq, AceView tries to represent all good quality cDNA sequences, even if many alternative mRNA forms become evident when this principle is applied. As a result, AceView represents the transcriptome in much greater depth than the other gene annotations, and it also houses a greater percentage of the probes designed by the microarray designers.

Please email us if one array you like is not in our current list and the probe sequences are available: we will remap it to the main annotations and place the results on our public ftp “downloads” site.

 

All the files posted on this ftp site since 2001 are still available.

Also see the GOLD and MAQC works.

April 2007

The Human Apr07 release aligns more than 7 million cDNA sequences (available March 26, 2007 in GenBank/ dbEST/ RefSeq) into genes on the human genome assembly NCBI_36/hg18 of March 2006.
The files available are:

• The coordinates of exons/introns, CDS and UTR of each mRNA in gff format, for non-cloud genes (26.6 MB)
• The mRNA sequences (including UTR parts) for the genes, known or unknown, but excluding the clouds, in fasta format (79.3 MB) (see the definition of clouds in the FAQ)
• The amino acid sequences of the best protein from each transcript, provided they look like ‘good proteins’, in fasta format (15.6 MB) Warning: like most proteins in SwissProt or RefSeq, these are conceptual translation products, most have not been observed experimentally.
• All AceView mRNAs, with no restriction: a comprehensive non redundant curated representation of all data submitted as cDNA sequences to GenBank and dbEST as of March 2007, in fasta format (119.3 MB).
• All peptides and proteins, with no restriction, for decoding mass spectrometry data, in fasta format (44.9 MB)
• The highly significant PFAM hits (position on genome of the corresponding gene, title and accession of the PFAM, coordinates of the domain in the proteins (3.4 MB)
• Added March 6, 2008: Mapping of human expression microarray probes Affymetrix HG-U133 Plus 2.0 GeneChip, Agilent Whole Human Genome Oligo Microarray, G4112A and Illumina Human-6 BeadChip, 48K v1.0 platforms is provided (53 MB, see this help). The AceView mapping should improve significantly the information you get from your arrays… Mapping array probes to genes in a systematic fashion is essential for comparing expression results obtained on different platforms, as well as to interpret results from a single platform (for example see the MAQC project). At the request of users, we make public the mapping of probes to AceView transcripts and the genome for some of the most used commercial microarray platforms.: in the first release (March 6, 2008), we provide alignments for the Illumina Human-6 BeadChip (48k v1.0), Agilent WHG (G4112A0) and Affymetrix HG-U133 Plus 2.0 GeneChip expression arrays, used in MAQC.
  The mapping is performed on the current human genome (build 36) as well as on three annotated transcripts sets, two from NCBI and one from EBI. We used all AceView transcripts (current release April 2007), all RefSeq (NM and XM, but not including the few non coding NR/XR) (downloaded January 2008), or all Ensembl models (downloaded January 2008).
  We use our program ProbeAlign, set to allow up to 2 mismatches (base difference or indel) for Agilent 60-mer probes and Illumina 50-mer, and up to 1 mismatch for Affymetrix 25-mer probes. For help on formats, see here. Also see the MAQC interesting outcome here, and the explanations here.
• Added February 8, 2009: The properties, sequence and support of all exon-exon junctions seen in mRNA sequences from any the NCBI sequence databases (GenBank, dbEST, Trace), as summarized in AceView. For help on content and format, see the 'File Format' tab.
  The list made public on February 8 2009 includes 368,512 exon junctions found in AceView, from mapping all cDNA sequences present in the NCBI database in April 2007. The list includes in particular the 199,952 introns in RefSeq NM/NR/XM/XR from November 22, 2008.
  The intron list will grow as more junctions are discovered, in particular in the deep sequencing data. Please tell us if you wish to be notified by mail when we update the list.
  As of today, there is support, in the small set of deep transcriptome data from SRA that we analyzed, for 173,000 junctions from RefSeq (86%) and 259,000 from AceView (68%). This zip includes two files:
  1- The AceViewRefSeqExonsJunctions.txt file provides the sequence of the junctions and their main properties (such as associated gene and transcripts, coordinates on the genome and boundaries of the intervening sequence).
  2- The complementary AceViewExonsJunctionsSupport.txt file provides the clones and accessions in GenBank/dbEST/Trace (but not SRA!) exactly supporting the junctions. Our criteria for a cDNA to “support a junction” is that the cDNA sequence exactly matches 16 bp centered on the junction (8 bases in each exon bordering the intron).
  There are currently 2,284,821 clones in the public databases supporting 367,750 junctions. Our aim is to annotate the deep transcriptome sequencing results on the web in our next release. We welcome wholeheartedly any new transcriptome data that you might want to see integrated: just send it to GEO or SRA!
• Added at the request of a user: disease related genes, according to OMIM, GAD and hints from Mesh/PubMed hunting we did

 

MAQC Microarray quality control project

From our analyses for the MAQC study, led by Leming Shi from FDA, we extracted these interesting data:

• A list of the 276,184 MAQC validated probes from all MAQC platforms, according to our unpublished analysis (File: ConfirmedTitratingProbes.txt). These probes map uniquely to a single gene and do not cross-hybridize. They have two desirable properties when hybridized to mixes of RNA samples A (Universal RNA Stratagene) and B (Ambion brain): they are “titrating”, i.e. signals for the 4 samples containing mixes of the two RNAs:  A (A100:B0), C (A75:B25), D (A25:B75), B (A0:B100) are in proper order, either monotonously increasing or decreasing when we average the normalized signals over the 15 replicas (minus a few outlier arrays). In addition, the direction of the differential expression for the gene measured agrees with at least 2 other platforms: A/C/D/B signals vary in the same direction in a majority of platforms assaying the gene, so in this sense, all probes in these files coherently and sensitively measure the validated differential expression A>B or B>A for the genes. Probes are ordered alphabetically, ABI, AFX (Affymetrix), AGL (Agilent), EPN (Eppendorf), GEH (General Electric Healthcare), GEX, ILM (Illumina), NCI (Operon), QGN (QuantiGene) and TAQ (TaqMan) (Note that this is a minimal list of validated probes, as some probes on the array might not have been testable by the two RNA samples selected for MAQC. But at least these probes have been proven to be good, sensitive and reliable.)
• A file giving a comparable measure of the melting temperature for all probes (files ending in .Tm).
• The mapping in AceView of all the MAQC probes to their desired (and undesired) target genes, including the identification of the specific alternative transcripts targeted, is available as a zip or a tar.gz compressed file. This file was generated as part of the Micro Array Quality Control project (MAQC study), we used ProbeAlign to map all probes sequences to the human genes (from the previous version of AceView, April 2005), including to their putatively cross-hybridizing targets. The total number of genes tested with gene specific probes by each genome wide array platform participating to the project, i.e. 29,040 genes for Affymetrix, 22,106 for Agilent, 21,943 for Applied Biosystems, 35,392 for Codelink General Electric, 27,463 for Illumina and 19,025 for the NCI/Operon array, are detailed in the Supplementary data, page 14, on the Nature Biotechnology website. 
  We hope this information may be useful, do not hesitate to feed us back on what your needs or questions are.

GOLD: Genomewide Optimization of Locus Description

The GOLD in-depth analysis we made in 2005 is still very actual: it includes a comparison between various cDNA aligners, including gnomon/splign and the EBI aligner, and the discovery of a multitude of interesting facts about the human genes. Results can be found in our (unpublished) article 'GOLD', which comes with a thorough supplementary material connecting on complete and detailed annotations. We are still now very proud of this work, which was investigating the reliability of cDNA to genome aligners (and was probably resented as too critical by the referee), but also summarizes the properties of transcription in human.

August 2005
The Human August 2005 AceView annotations were performed on the human genome NCBI_35/hg17 of July 2004.
This release aligns ESTs, mRNAs and RefSeqs in GenBank or dbEST on September 24 on the human genome build, NCBI_35 /hg17 of July 2004.
Two files were archived in July 07, the files still available are:
 The coordinates of exons/introns, CDS and UTR of each mRNA in gff format, for non-cloud genes (26.8 MB)
The amino acid sequences of the best and good CDSs or ORFs in non-cloud mRNA, in fasta format (21.9 MB)
All peptides and proteins, with no restriction, for decoding mass spectrometry data, in fasta format (33.9 MB)
The mRNA sequences (including UTR parts) for all non-cloud genes, in fasta format (77.8 MB), and the cloud
The highly significant PFAM hits (position on genome of the corresponding gene, title and accession of the PFAM, coordinates of the domain in the proteins (3.2 MB)

December 2003
The Human December 2003 (Dec03) AceView annotations were performed on the human genome NCBI_34/hg16 of July 2003.
The amino acid sequences of each CDS or ORF in fasta format (16.3 MB)
The coordinates of exons/introns, CDS and UTR of each mRNA in gff format (18.5 MB)
The mRNA sequences (including UTR parts) in fasta format (77.9 MB)
The highly significant PFAM hits (position on genome of the corresponding gene, title and accession of the PFAM (730 kB)

May 2003
The human AceView annotations were performed on the previous human genome build, NCBI_33 genome build of May 2003.
The amino acid sequences of each CDS or ORF in fasta format (14.5 MB)
The coordinates of exons/introns, CDS and UTR of each mRNA in gff format (20.5 MB)
The mRNA sequences (including UTR parts) in fasta format (74.6 MB)
The highly significant PFAM hits (position on genome of the corresponding gene, title and accession of the PFAM (604 kB)

December 2002
The human AceView annotations were performed on the previous human genome build, NCBI_31 genome build of Decembre 2002.
The amino acid sequences of each CDS or ORF in fasta format (14.6 MB)
The coordinates of exons/introns, CDS and UTR of each mRNA in gff format (10.5 MB)
The mRNA sequences (including UTR parts) in fasta format (65.6 MB)
The highly significant PFAM hits (position on genome of the corresponding gene, title and accession of the PFAM (582k)

August 2002
The human AceView annotations were performed on the human genome build, NCBI_30 genome build of August 25 2002.
The amino acid sequences of each CDS or ORF in fasta format (11.5 MB)
The coordinates of exons/introns, CDS and UTR of each mRNA in gff format (12.9 MB).
The mRNA sequences (including UTR parts) in fasta format (46.8 MB)
Please quote us as D,J and Y Thierry-Mieg, M.Potdevin, M.Sienkiewicz, V.Simonyan, www.humangenes.org: Construction and automatic annotation of cDNA-supported genes using Acembly, unpublished

May 2002
The human AceView annotations were performed on the human genome build, NCBI_29 of may 25 2002:
The amino acid sequences of each CDS or ORF in fasta format (12.0 MB)
The coordinates of exons/introns, CDS and UTR of each mRNA in gff format (15.3 MB)
Please quote us as D,J and Y Thierry-Mieg, M.Potdevin, V.Simonyan, www.humangenes.org: Construction and automatic annotation of cDNA-supported genes using Acembly, unpublished

January 2002
The human AceView annotations were performed on the human genome build, NCBI_28 of January 2002:
The amino acid sequences of each CDS or ORF in fasta format (9.5 MB)
The mRNA sequences (including UTR parts) in fasta format (40.8 MB)
The coordinates of exons/introns, CDS and UTR of each mRNA in gff format (9.5 MB)

November 2001

The human AceView annotations were performed on the human genome build, NCBI_28 of January 2002:
The amino acid sequences of each CDS or ORF in fasta format (10 MB)
The mRNA sequences (including UTR parts) in fasta format (42.8 MB)
The coordinates of exons/introns, CDS and UTR of each mRNA in gff format (10 MB)

Mouse June 2007

The Mouse June 2007 AceView release aligns more than 3 million cDNA sequences (available June 22, 2007) into genes and transcripts on the Mus musculus NCBI genome 37 (June 2007). The files available are:

• The coordinates of exons/introns, CDS and UTR of each mRNA in gff format, for non-cloud genes (17.4 MB)
• The mRNA sequences (including UTR parts) for the genes, known or unknown, but excluding the clouds, in fasta format (60.5 MB) (see the definition of clouds in the FAQ)
• The amino acid sequences of the best protein from each transcript, provided they look like ‘good proteins’, in fasta format (12.2 MB) Warning: like most proteins in SwissProt or RefSeq, these are conceptual translation products, most have not been observed experimentally.
• All AceView mRNAs, with no restriction: a comprehensive non redundant curated representation of all data submitted as cDNA sequences to GenBank and dbEST as of March 2007, in fasta format (81.3 MB).
• All peptides and proteins, with no restriction, for decoding mass spectrometry data, in fasta format (28.9 MB)

Content 4

Content 2