SFESA (Shift to Fix secondary structure ElementS in Alignments) is a tool to refine pairwise protein sequence alignment, with a combination of sequence and structural scoring by locally shifting secondary structure elements.
Overview of alignment refinement process
SFESA is a method that can refine an original alignment (for example, PROMALS [1] pairwise alignment) by evaluating local shifts of template-defined secondary structural elements based on a novel scoring function (see flowchart). Firstly, a pairwise alignment input is treated as original alignment. Otherwise SFESA generates PROMALS alignment as the original alignment when two unaligned sequences as input. Secondly, the template sequence is subject to PSI-BLAST [2] searches to retrieve closest structure from our structural representative database if no template structure is given from input. Then, if the sequence identity between structure file and template sequence in input is above the threshold (default is 50%), the input of original alignment and template structure is fed into alignment refinement pipeline. Otherwise, only original alignment will be return without refinement. There are four modes for alignment refinement method: SFESA_O, SFESA_O+G, SFESA_O+G+M, SFESA_O+G+M+S depending on different gap processing methods and contact-based energy matrices (See parameters). Finally, the original alignment and the refined alignment will be shown as result.
SFESA input
Two unaligned sequences (or a pairwise alignment) and/or a structure of template can be input to SFESA. Strongly recommend to provide a structure of template to increase alignment quality.
Input sequences of query and template: (1). A pairwise alignment in FASTA format, or (2). two unaligned sequences in FASTA format.Input sequences should be in FASTA format: Each sequence record consists of a description line followed by line(s) of the sequence.
Note: The first character of each description line should be a greater-than (">") sign.
(1). Input of a pairwise alignment in FASTA format.
For this input format, SFESA will make local shifts of secondary structure elements starting from this input alignment.
Example:
>d1ja1a3 (query) RLP-FKSTTPVIMVGPGTGIAPFMGFIQERAWLREQGKEVGETLLYYGCRRSDEDYLYREELARFHKDGALTQLNVAFSREQAHKVYVQHLLKRDREHLWKLIHEGGAHIYVAGDARNMAKDVQNTFYDIVAEFGPMEHTQAVDYVKKLMTKGRYSLNVWS >d2piaa2 (template) EFPLDKRAKSFILVAGGIGITPMLSMARQLRAEG-----LRSFRLYYLTRDP-EGTAFFDELTSDEWRS-DVKIHHDHGDPT-KAFDFWSVFEKS---------KPAQHVYCCGP--------QALMDTVRDMTGHWPSGTV-------------HFE---(2). Input of two unaligned sequences in FASTA format.
For this input format, SFESA will generate a pairwise alignment for the two sequences by PROMALS,
then make local shifts of secondary structure elements.
Note: Please click the choice of "Two unaligned sequences" above the input frame, otherwise the default is to use input as an pairwise alignment.
Example:
>d1ja1a3 (query) RLPFKSTTPVIMVGPGTGIAPFMGFIQERAWLREQGKEVGETLLYYGCRRSDEDYLYREELARFHKDGALTQLNVAFSREQAHKVYVQHLLKRDREHLWKLIHEGGAHIYVAGDARNMAKDVQNTFYDIVAEFGPMEHTQAVDYVKKLMTKGRYSLNVWS >d2piaa2 (template) EFPLDKRAKSFILVAGGIGITPMLSMARQLRAEGLRSFRLYYLTRDPEGTAFFDELTSDEWRSDVKIHHDHGDPTKAFDFWSVFEKSKPAQHVYCCGPQALMDTVRDMTGHWPSGTVHFE
Any non-alphabetical character in the input sequences is ignored by SFESA.
A note for sequence names: Certain characters in sequence names are changed to "_", including space, tab, and *?'`"&|\/{}()[]$; (.(dot) and - are kept).
Input structures: It is recomended to upload an structure file for template or template homolog. The uploading structure files should be in PDB format. If not provided, search agaisnt our database to find the cloest homolog structure for the template. The sequence that is found to be closest to the provided structure or the structure database is assigned as the Template. The other sequence is assigned as the Query. Alignment will be refined only when the sequence identity between structure and template sequence (idential residues devied by all aligned positions) is above the threshold (default is 50%).
Input of template structure example (part) (see complete structure, click here):
ATOM 795 N GLU A 1 8.804 32.358 19.734 1.00 12.73 N ATOM 796 CA GLU A 1 9.685 32.086 18.600 1.00 14.50 C ATOM 797 C GLU A 1 10.448 33.301 18.068 1.00 15.83 C ATOM 798 O GLU A 1 11.330 33.155 17.221 1.00 16.89 O ATOM 799 CB GLU A 1 8.880 31.462 17.446 1.00 17.54 C ATOM 800 CG GLU A 1 8.258 30.101 17.753 1.00 16.58 C ATOM 801 CD GLU A 1 7.417 29.543 16.614 1.00 16.05 C ATOM 802 OE1 GLU A 1 6.651 30.296 15.981 1.00 20.00 O ATOM 803 OE2 GLU A 1 7.468 28.325 16.391 1.00 18.78 O ATOM 804 N PHE A 2 10.062 34.497 18.492 1.00 13.99 N ATOM 805 CA PHE A 2 10.683 35.714 17.982 1.00 13.09 C ATOM 806 C PHE A 2 10.788 36.758 19.102 1.00 13.51 C ATOM 807 O PHE A 2 10.157 37.807 19.040 1.00 15.07 O ATOM 808 CB PHE A 2 9.859 36.249 16.806 1.00 13.31 C ATOM 809 CG PHE A 2 10.557 37.286 15.962 1.00 11.78 C ATOM 810 CD1 PHE A 2 11.882 37.124 15.579 1.00 12.00 C ATOM 811 CD2 PHE A 2 9.856 38.378 15.473 1.00 12.73 C ATOM 812 CE1 PHE A 2 12.486 38.033 14.712 1.00 15.85 C ATOM 813 CE2 PHE A 2 10.452 39.286 14.611 1.00 14.41 C ATOM 814 CZ PHE A 2 11.765 39.114 14.228 1.00 11.11 C ATOM 815 N PRO A 3 11.539 36.447 20.176 1.00 16.07 N ATOM 816 CA PRO A 3 11.504 37.278 21.381 1.00 12.77 C ATOM 817 C PRO A 3 12.422 38.484 21.330 1.00 13.95 C ATOM 818 O PRO A 3 13.440 38.486 20.635 1.00 14.16 O ATOM 819 CB PRO A 3 11.925 36.315 22.485 1.00 15.51 C ATOM 820 CG PRO A 3 12.873 35.417 21.805 1.00 14.27 C ATOM 821 CD PRO A 3 12.264 35.186 20.429 1.00 13.84 C ATOM 822 N LEU A 4 11.972 39.548 21.985 1.00 13.07 N ATOM 823 CA LEU A 4 12.773 40.728 22.293 1.00 14.75 C ATOM 824 C LEU A 4 14.026 40.348 23.073 1.00 17.28 C ATOM 825 O LEU A 4 13.915 39.740 24.129 1.00 16.56 O ATOM 826 CB LEU A 4 11.947 41.662 23.163 1.00 18.57 C ATOM 827 CG LEU A 4 11.437 42.989 22.645 1.00 21.68 C ATOM 828 CD1 LEU A 4 11.549 43.062 21.139 1.00 24.97 C ATOM 829 CD2 LEU A 4 10.010 43.146 23.114 1.00 21.31 C ATOM 830 N ASP A 5 15.198 40.802 22.629 1.00 16.86 N ATOM 831 CA ASP A 5 16.432 40.581 23.384 1.00 17.70 C ATOM 832 C ASP A 5 16.420 41.250 24.772 1.00 18.70 C ATOM 833 O ASP A 5 16.162 42.443 24.876 1.00 19.51 O ATOM 834 CB ASP A 5 17.623 41.095 22.590 1.00 21.09 C ATOM 835 CG ASP A 5 18.938 40.534 23.084 1.00 21.43 C ATOM 836 OD1 ASP A 5 19.512 41.116 24.021 1.00 23.83 O ATOM 837 OD2 ASP A 5 19.407 39.536 22.504 1.00 27.41 O
OR enter PDB identifier and chain identifier: Input of 4-digital formatted PDB identifier (e.g.: 1ja1) and chain identifier can directly load the strucutre from PDB Database. If no chain identifier is given, the whole chain will be used.
Input Email: SFESA jobs usually take minutes to hours for sequence alignment. It takes longer if seuqnece with long length or many homologs. Thus, it is highly recommended that an email address is provided so that the link to your result is sent to you when the alignment is finished.
Input job name: Assign your sequences a short name can help identify your alignment job. This name will appear in the subject line of the email sent to you.
SFESA output
SFESA web server provides results include the following information.
1. Original alignment with secondary structure and colored alignment blocks (based on secondary structure of template).
The first line in this block shows predicted secondary structure from PSIPRED [3] for query sequence. The last two lines show the secondary structure from DSSP [4] and PALSSE [5] ("H"-Helix, "S"-Strand and "C"- Coil). The template elements are based on PALSSE [5]. The "Query" and "template" are from the starting sequence alignment (PROMALS pairwise alignment when just two unaligned sequences as input or the input alignment). The line started with "Number" above query or below template alignment shows the position number of the residue below and above in query or template.
If the sequence identity between of structure file and template sequence in input (idential residues devied by all aligned positions) is below the threshold (default is 50%), only this original alignment part is generated as final result.
The "Helix" alignment blocks are shown alternately in Red and Orange in alignment.
The "Strand" alignment blocks are shown alternately in Blue and Darkgreen in alignment.
Starting alignment example:
Original Pairwise Alignment (no SFESA refinement)
2. Refined alignment with secondary structure and colored alignment blocks (based on secondary structure of template).
If the sequence identity between of structure file and template sequence in input (idential residues devied by all aligned positions) is above the threshold (default is 50%), the following parts including such refined alignment and shifting details will be generated.
The line started with "Cm1" and "Cm2" represent the comparison of refined and original alignment. "Cm1" shows the sign of the query residue shifting ("+": query residue shifted towards C-terminal; "-": query residue shifted towards N-terminal;)) while "Cm2" shows the query residue shifting number. If the query residue is aligned to a gap in both original and refined alignment, "Cm1" leaves blank and "Cm2" shows "-". If query residue is aligned to one residue in original alignment but aligned to gap in refined alignment, "Cm1" leaves blank and "Cm2" shows "*". If template residue is aligned to gap, both "Cm1" and "Cm2" leaves blank. Other lines are same as above original alignment explanation.
The "Helix" alignment blocks are shown alternately in Red and Orange in alignment.
The "Strand" alignment blocks are shown alternately in Blue and Darkgreen in alignment.
In SFESA alignment, the alignment blocks shifted are marked with Underscore.
Refined alignment example:
Refined Alignment by SFESA (SFESA_O+G+M)
3. List of all alignment blocks in SFESA.
This part shows a table containing all alignment blocks (based on secondary structure of template) shifting results.
The 1st column is the number of alignment block. Clicking on this number will direct to the details of every shifting position of this alignment block.
The 2nd and 3rd column is the template secondary structure element starting and ending residue position number of this alignment block.
The 4th column shows the secondary structure type.
The 5th column shows the original alignment block.
The 6th column indicates the shifting result of this alignment block. If shifted, a format of "Gap mode [shifting number]" is showed. There are three gap modes: Original (no change of original alignment block), Left (residues in alignment blocks can be aligned to leftmost while all gaps are put to the opposite side before shifting) and Right (residues in alignment blocks can be aligned to rightmost while all gaps are put to the opposite side before shifting). And the default setting for shifting number is from -4 to +4.
The 7th column shows the modified alignment block if the alignment block is shifting. Otherwise the 7th column shows "-" that means no change.
The colored line means this alignment block is shifted in refined alignment.
Example:
List of all elements in SFESA
| Alignment Block Number | Template Start | Template End | Secondary Structure Type | Original Alignment Block | Shift by SFESA (Gap Mode [Shifting Number]) | Refined Alignment Block |
| 1 | KSFILVAG |
- | ||||
| 2 | GIGITPMLSMARQLRAEG |
- | ||||
| 3 | RSFRLYYLTRD |
- | ||||
| 4 | DVKIHHDH |
DVKIHHDH |
||||
| 5 | DFWSVFEKS |
- | ||||
| 6 | AQHVYCCG |
- | ||||
| 7 | P--------QALMDTVRDMTG |
PQALMDTVRDMTG |
4. Shifting details of all elements in SFESA
This part shows tables for each alignment blocks which contains all possible alignment block variants (alignment blocks without gaps: 1 original + 8 variants; alignment blocks with gaps: 1 original + 9 "left" variants + 9 "right" variants).
The 1st column in each table is gap mode. There are three gap modes if there are gaps in this alignment block: Original (no change of original alignment block), Left (residues in alignment blocks can be aligned to leftmost while all gaps are put to the opposite side before shifting) and Right (residues in alignment blocks can be aligned to rightmost while all gaps are put to the opposite side before shifting).
The 2nd column is the shifting number. The default setting for shifting number is from -4 to +4.
The 3rd column indicates if such variant is unique or the same as one shown previously. If not unique, the previous shifting result is shown.
The 4th column shows the alignment block variant with extended residues in two ends. The residues in blue and pink show the original alignment block in query and template. The maximal boundaries of marked original alignment block defined the boundaries of the alignment variant.
The 5th, 6th, 7th and 8th column show the sequence score, structure score, combined score 1 and combined score 2 of such alignment block variant.
The line in red means this alignment block variant is the final choice in refined alignment.
Example:
Scoring details of shifts in alignment block number 4:
| Gap Mode | Shift Number | Unique? | Alignment Variants | Sequence Score | Structure Score | Combined Score I | Combined Score II |
-----DVKIHHDHGDPT- |
|||||||
S----DVKIHHDHGDPT- |
|||||||
RS---DVKIHHDHGDPT- |
|||||||
WRS--DVKIHHDHGDPT- |
|||||||
EWRS-DVKIHHDHGDPT- |
|||||||
DEWRSDVKIHHDH-GDPT |
|||||||
DEWRSDVKIHHDH--GDP |
|||||||
DEWRSDVKIHHDH---GD |
|||||||
DEWRSDVKIHHDH----G |
Alignment parameters are listed below
SFESA refined alignment mode
When we analyzed PROMLAS alignments, we found that secondary structure elements are often misaligned in alignment generated by PROMALS by a few residues. And the main reason is PROMLAS cannot introduce gaps in order to avoid the gap penalties. In these cases correct alignment solutions can be found within a limited set of local shifts of secondary structure elements. Data shows about 80% of the improvable alignment block can be refined by up to 4 residue local shifts. Thus, the alignment variants generated for one alignment block is limited to +/-4 shifting and the number of variants is up to 8.
Another scenario is that there are gaps in query or template alignment blocks. Above +/-4 shifting works for this scenario. In addition, there are more alignment variant options if the pre-process of gaps is done. It is known that most secondary structure elements should be aligned to secondary structure elements without gaps inserted, thus we can process the gaps by putting them to one side without interrupting the secondary structure elements. Residues in alignment blocks can be aligned to leftmost or rightmost while all gaps are put to the opposite side.These two alignment variants can be considered as new original alignment blocks, so 8 alignment variants can be produced for each new origin. In this case, computational experiments show that it's better to consider only gap-processed alignment variants. That is to say for these cases with gaps, SFESA can generate up to 18 (1+1+8+1+8) alignment variants.
On the other hand, contacted residues are defined as residue pair within a distance cutoff between them and the contacting network among the protein supports the global protein structure. In the template of one alignment, the contacting network can be identified based on the template known structure. One residue in the template cannot only contact with sequence neighboring residues but also contact with the structurally closed residues far ways in sequence order. These contacts defines residues structural environment in the template, and the correctly aligned equivalent residues in the query should have similar structural environment and vice versa. Based on the deduced residue-residue contact energy matrix by Miyazawa and Jernigan [6], the total contact energies (structure score) of query residues in the alignment block can be utilized to select better alignment variant.
Our study shows contact matrix derived from alignment and different definition of residue-residue contact can help to better differentiate correct and wrong alignment. The new contact cutoff is 6.5A between any atoms of two residues. Alignments for deriving are original PROMALS alignments from training dataset. Each alignment block is allowed to shift +/-4 residues (8 variants), and then the best one is selected by DALI-dependent accuracy [7] and the other 8 alignment variant (or original alignment block) are considered as decreased cases.
SFESA (O) means to use up to 8 variants and Miyazawa-Jernigan contact matrix;
SFESA (O+G), considering gap processing, used up to 18 variants; besides this;
SFESA (O+G+M) tried our derived contact matrix;
SFESA (O+G+M+S) used SSVM in second filter instead of Scomb_II.
The default mode is SFESA (O+G+M) which has the highest overall alignment accuracy in our inhouse data. But, other three modes show a better result in some cases.
Structure sequence identity threshold (Identity threshold above which structure of template is applied)
The parameter "structure sequence identity threshold" is the sequence identity threshold between template structure (the input structure or the closest homolog structure searching against our structure database if no structure input) and template sequnece in input. The sequence identity is idential residues devied by all aligned positions (ignore resides position aligned to gaps).
Our alignment refinement method replies on the contacts from the template structure. More accurate template structure given, the higher alignment quality generated. And the low sccurate structure may hurt the original alignment quality. Thus, a accurate structure of template is strongly recommended to upload.
The default value is 0.5 (range is 0-1). This means that original alignment will be refined by our method only if at least 50% of aligned residues between template structure and sequence are identical.
Maximal residue positions to shift
After secondary structure elements are recognized from template structure, all alignment blocks are generated. For each alignment block, our method can locally shift to generate several alignment block variants. Then, a better (or the original) alignment block variant can be selected based on a combination of sequence and structural scoring. Thus, it is very important to choose the maximal shifting positions. The more shifting positions tried, the higher probability the better variant generated but the more difficult the better correctly selected. Our computational result shows that 4 residue shifting is the most efficient one which can generate 80% better alignment blocks in our data.
The default setting is 4 (range is integer above 1).
However, this parameter can be customized to check different alignment block variants with different maximal shifting position number. For example, n maximal residue positions means 2n alignment block variants if no gaps processing while 4n+2 alignment block variant if gaps processing considered.
Non-gap threshold above which elements in template is applied SFESA
The secondary structure elements are recognized from template structure. It's probable that such elements are insertions if most residues of them are aligned to gaps in query and template alignment. It is meaningless to fix alignment blocks recognized from such insertion elements. Thus, a threshold parameter is to prevent such alignment block processing.
The default value is 0.5 (range is [0-1)). This means the alignment block is applied SFESA to refine only if the non-gap aligned residue percentage is more than 50% in the alignment block recognized from template secondary
structure elements.
Automatically assign structure for sequence in input
The default setting is "Yes". That means, the sequence that is found to be closest to the provided structure or the structure database is assigned as the Template (T). The other sequence is assigned as the Query (Q). If "No" is selected, the input user provided is assumed to put query sequence firstly followed by template sequence (the structure is for the second sequence in input).
Parameters for generating Psi-blast profile
Iterations
maximum number of iterations done by PSI-BLAST. The range is integer between 1 and 8.
E-value
only hits with an e-value less than this threshold will be included in the next iteration.
Identity cutoff below which distant homologs are removed
any PSI-BLAST hit with a sequence identity to the query less than this cutoff will be removed. Divergent homologs could negatively affect sequence profile quality. (default: 0.2, corresponding to 20% sequence identity. The range is 0-1.)
References
1. Pei, J. and Grishin, N.V. (2007) PROMALS: towards accurate multiple sequence alignments of distantly related proteins. Bioinformatics, 23, 802-808.
2. Altschul S.F., Madden T.L., Schaffer A.A., Zhang J., Zhang Z., Miller W., Lipman D.J. (2005) Gapped BLAST and PSI-BLAST: a new generation of protein database search programs. Nucleic Acids Res 25:3389-3402.
3. Jones D.T. (1999) Protein secondary structure prediction based on position-specific scoring matrices. J Mol Biol , 292:195-202.
4. Kabsch, W. and Sander, C. (1983) Dictionary of protein secondary structure: pattern recognition of hydrogen-bonded and geometrical features. Biopolymers, 22, 2577-2637.
5. I.Majumdar, S.S.Krishna and N.V.Grishin (2005) PALSSE: A program to delineate linear secondary structural elements from protein structures. BMC Bioinformatics 6: 202.
6. Miyazawa, S. and Jernigan, R.L. (1999) An empirical energy potential with a reference state for protein fold and sequence recognition. Proteins, 36, 357-369.
7. Holm, L. and Sander, C. (1996) Mapping the protein universe. Science, 273, 595-603.