Seq2Ref - Retrieve reference proteins in BLAST result
General Description and submission:
Seq2Ref is a server to facilitate the functional interpretation of a protein. It performs 
BLAST search for your query and retrieves the reference proteins, defined as the experimentally 
studied or manually curated proteins, from three resources, i.e. NCBI, PDB and Swiss-Prot. 
According to the results from reciprocal BLAST, Multiple Sequence Comparison(MSC) and 
pairwise sequence comparison, we assign a rating score to the reference protein to indicate 
homologous relationship to the query. 

Submission:
To submit a job to Seq2Ref, you need to provide your protein and your email address; 
Three formats, i.e. FASTA format, plain-text format and gi number, of your protein 
are accepable for your query protein. Generally, the total running time is 5--10 min 
for fast mode and about 1 hour for slow mode. Once the job is done, an email containing 
the link to the result page will be sent to you. 

Terms in the input interface:

1. Input email: Accidentally closing the webpage would cause the unexpected lost track 
of the job; moreover, the job in slow mode can take a relatively long time (approximately 
1h) to complete. To keep track of the job, an email address is required so that the link 
to your result is sent to you when the job is done.

2. Input job name: Assign a short name can help identify your Seq2Ref job. This name 
will appear in the subject line of the email sent to you. Input job name is not required.

3. BLAST parameters: those are the parameters used in BLAST search.
    3.1 Expect threshold or e-value: Expected number of chance matches in a random model. 
Smaller e-value indicates better homologous relationship. Detailed explanation refers to: 
http://www.ncbi.nlm.nih.gov/BLAST/blastcgihelp.shtml#expect
    3.2 Iteration: the number of iterations performed in the Position-Specific Iterated BLAST 
(PSI-BLAST). If iteration equals 1, BLAST instead of PSI-BLAST will be performed.
    3.3 Max target sequences: Maximum number of aligned sequences shown in BLAST result and 
to be considered in detecting reference proteins.
    
4. Fetching pubmed literature: the retrieval of reference proteins linked to pubmed 
literature requires inquiring the cross-links of the databases maintained by NCBI. Two 
modes, i.e. the fast mode and slow mode are developed to analyze the cross-links for 
pubmed literature retrieval.
    4.1 Fast mode: In fast mode, reference proteins and pubmed articles are retrieved by 
searching a pre-processed local database (updated every six months) consisting of all 
the reference proteins in NCBI databases. With default parameters, it takes 5--10 min 
to complete a job. Fast mode fetch ensures the high speed of the server, but takes the 
risk of missing pubmed articles recently updated by NCBI.
    4.2 Slow mode: In slow mode, reference proteins and pubmed articles are retrieved by 
real-time consulting NCBI Entrez. With default parameters, it takes about 1h to complete 
a job. Slow mode fetch ensure the most updated set of reference proteins and pubmed articles, 
at the cost of the long running time of the server. 

Explanation of results:
The results are made of query related information and reference protein information:

1. Query related information: the information related to your query sequence. It contains 
the first hit in BLAST result, the input sequence, the plain-text BLAST result, and the 
genome of the query sequence (if detected). 

2. Reference protein information: information regarding reference proteins from three sources, 
i.e. NCBI, PDB, Swiss-Prot. The summary table summarizes the key parameters and descriptions of 
the reference proteins; detailed descriptions present all the parameters, sequence alignment and 
the related information, such as publications, to interpret the protein function. In the detailed 
description, proteins with rating score lower than 3 are hidden by default. 


Terms in the output interface:

1. Score: the rating score evaluating the homologous relationship between the query and the 
hit. Three different aspects, i.e. reciprocal BLAST, multiple sequence comparison and pairwise 
comparison, are considered. For score assignment please refer to the rating system.

2. Expect threshold or e-value: Expected number of chance matches in a random model. 
Smaller e-value indicates better homologous relationship. Detailed explanation refers to: 
http://www.ncbi.nlm.nih.gov/BLAST/blastcgihelp.shtml#expect

3. Identities: The extent to which two sequences have the same residues at the same 
positions in an alignment, often expressed as a percentage. It is calculated by taking the 
ratio of the identical positions to the total aligned positions in the alignment.

4. Coverage: The extent to which the aligned sequence region covers the full sequence length in 
an alignment, often expressed as a percentage. It is calculated by taking the ratio of the non-gap 
aligned region to the longest protein sequence in the alignment. 
    
5. Iteration: the number of iterations performed in the Position-Specific Iterated BLAST 
(PSI-BLAST). If iteration equals 1, BLAST instead of PSI-BLAST will be performed. Iteration in the 
summary table will be hidden if only BLAST is performed.

6. Pubmed: the pubmed article(s) linked to the protein by NCBI.

7. The evidence code: The evidence codes indicate the observations used to calculate the rating 
score. The codes have three aspects. 
    7.1 The first code, which consists of two components separated by the slash, denotes the 
observations in reciprocal BLAST. The first component is the result of the BLAST search against 
the genome of the hit (forward direction) and can be 'yes', 'no' and 'NA'; the second component 
is the result of the BLAST search against the genome of the query (backward direction) and can 
be 'yes', 'no' and 'NA'. 'yes' and 'no' indicate whether or not the query/hit is the best hit; 
'NA' indicates the unavailability of the complete genome sequence of the query/hit. 
    7.2 The second code, formated as 'msc' plus the rating score assigned, indicates the 
observation in multiple sequence comparsion. For the score assignment please refer to the
rating system.
    7.3 The third code, formated as 'pw' plus the rating score assigned, indicates the 
observation in pairwise comparison. For the score assignment please refer to the rating system.  

8. Reference(s): the publications are sorted by dates. The format of publication is 
'authors: title; journal; pubmed id. Pubmed center id (if applicable)'. For the publication 
with more than 10 authors, only the first five and last five authors are listed.

9. Positives: The extent to which two sequences have the residues of similar properties 
at the same positions in an alignment, often expressed as a percentage. It is calculated by 
taking the ratio of positive positions that contain the positive-scored residue pair in 
BLOSUM matrix to the alignment length. 

10. Gaps: The extent to which two sequences are not aligned in an alignment, often expressed 
as the ratio of the unaligned positions to the alignment length. 

11. Functional annotation: the function described in the 'General annotation (Comments)' panel of 
the Swiss-Prot database record.




The rating system:
Homologous relationships are evaluated by the rating system with scores ranging from
1 to 6. Higher score indicates better homologous relationship; proteins with score 
lower than 3 would be considered remote homologs whose functions may have diverged.
Three aspects are considered as follows.
1. Reciprocal BLAST:
   Forward direction best	---------------- True=+1;   False=+0;   NA=+0.25;
   Backward direction best	---------------- True=+1;   False=+0;   NA=+0.25;
2. Accepted in MSC with coverage>80% and : 
   Identity cutoff 60%    	---------------- True=+0.5; False=+0;
   Identity cutoff 50% 		---------------- True=+0.5; False=+0;
   Identity cutoff 60%  	---------------- True=+0.5; False=+0;
3. Pairwise comparison to the query with coverage>80% and :
   Identity>60%      		---------------- True=+0.5;     False=+0;
   Identity>50%       		---------------- True=+0.5;     False=+0;
   Identity>40%        		---------------- True=+0.5;     False=+0;
Notes: 
1. 'Forward direction' means using the query to BLAST against the genome of the hit.
2. 'Backward direction' means using the hit to BLAST against the genome of the query.
3. 'NA' means no whole genome sequence is available.


Vocabulary:

1. Reference proteins: the proteins either with experimental studies or manually curated; 
Their function interpretations are supported by expermental data and not origniated from 
automatic annotation transfer.

2. BLOSUM (BLOcks of Amino Acid SUbstitution Matrix) matrix: The substitution matrix used 
for sequence alignment of proteins. Details refer to http://en.wikipedia.org/wiki/BLOSUM