2 Pairwise Sequence Alignment

Most sequence database search programs depend to some extent on pairwise sequence alignment - the process of identifying the mathematically optimal way to line up the symbols of two sequences in order so that they are as similar as possible.

Consider the two words COELACANTH and PELICAN. These words are obviously different, but they are also similar enough that we can shift their letters about to make an alignment between them. For instance, we might align the words as:

COELACANTH
-PELICAN--

so that the E, L, C, A, and N letters are in the same columns. The only differences are then the matching of the O of COELACANTH with the P of PELICAN, and the need to insert gaps (- characters) to indicate where there is no corresponding letter, in the shorter word.

Is there another - or a better - way to align these words?

What if we were to align the words COELACANTH and PELICAN differently, like this?

COELACANTH
P-ELICAN--

Is this a better or worse alignment of those two words than the one you originally saw? Why do you think this? What kind of rule would you devise to make a choice of which was the better alignment?

Figure 2.1: A pelican eyeing up a coelacanth for lunch

2.1 Pairwise Alignment is a Solved Problem

Pairwise sequence alignment is a solved mathematical problem. What that means is that, if we are given:

two sequences
a scoring scheme for matching symbols, such as:
- a match scores +1 (e.g. A and A are in the same column)
- a mismatch scores 0 (e.g. C and G are in the same column)

we can guarantee to find the alignment with the largest possible score, using an algorithm (i.e. a set of reproducible mathematical steps). The alignment with the largest score is considered the optimal alignment. Such algorithms find one of two kinds of optimal alignment: either a global or a local optimal alignment.

Is the optimal alignment biologically “real”?

Given that we can find an alignment with a maximum score - the optimal alignment - why should we believe that this is the best biological alignment?

If we change the scoring scheme we might change the alignment, so what features of a scoring scheme might make it a better or worse representation of biological “reality”?

Important

On this page you will carry out two pairwise alignments, one global and one local, of the same input sequences, to examine the differences between the two approaches. You will use the tools Needle and Water from the EMBOSS suite of sequence analysis tools (Madeira et al. (2022)).

2.2 Exercise 1: Global Pairwise Alignment

Global sequence alignment finds the optimal alignment acriss the full lengths of both sequences being aligned.

Note

Given a scoring scheme, the Needleman-Wunsch algorithm is guaranteed to find an optimal global alignment between two sequences.

Exercise 1 task

Using the two DNA sequences below, find their optimal global alignment using the EMBOSS Needle program at EMBL-EBI.

>sequence_1
ACTACTAGATTACTTACGGATCAGGTACTTTAGAGGCTTGCAACCA

>sequence_2
TCACTTTAGAGGCTTGCAACCAGTACTTTAGAGGCTTGCTCAC

2.2.1 Open the `EMBOSS Needle` webpage

Open the link https://www.ebi.ac.uk/jdispatcher/psa/emboss_needle in a web browser, to reach the landing page (Figure 2.8)

Figure 2.2: The `EMBOSS Needle` landing page.

2.2.2 Enter the input sequences

Paste the two sequences given above into the two sequence input boxes, as shown in Figure 2.3.

Figure 2.3: The `EMBOSS Needle` page with two input sequences provided.

2.2.3 Set appropriate parameters

The two sequences you have entered are DNA sequences, so check the radio button marked DNA (Figure 2.4)

Figure 2.4: `EMBOSS Needle` input parameter set to `DNA`.

2.2.4 Run the alignment

Click on the Submit button at the bottom of the page to run the alignment. You can give your job a title, if you like (Figure 2.5).

Figure 2.5: `EMBOSS Needle` submit button.

A holding window will appear while the job runs (Figure 2.6), and after a short wait you will be presented with the results page (Figure 2.7).

Figure 2.6: `EMBOSS Needle` holding window.

Figure 2.7: `EMBOSS Needle` results window.

2.2.5 Interpret the results

The key result returned by Needle is, of course, the overall sequence alignment, which is provided at the foot of the results page:

sequence_1         1 ACTACTAGATTACTTACGGATCAGGTACTTTAGAGGCTTGCAACCA----     46
                                         ||    ||||||||||||||||||||    
sequence_2         1 --------------------TC----ACTTTAGAGGCTTGCAACCAGTAC     26

sequence_1        47 -----------------     46
                                      
sequence_2        27 TTTAGAGGCTTGCTCAC     43

What do you notice about the alignment?

All bases from both sequences are represented in the alignment, even though the sequences are different lengths, and most bases are not found to align.
Some bases in one sequence are aligned to the same base in the other - an identity - as indicated by the pipe symbol (|)
Some bases in the upper sequence are not aligned to any bases in the lower sequence, and vice versa. Where this happens, the other sequence is padded out with gaps (-). In those regions, the runs of bases may be referred to as insertions, and the missing sections represented as gaps may be called deletions.

2.3 Exercise 2: Local Pairwise Alignment

Local sequence alignment finds the optimal (i.e. highest scoring) alignment between two subregions of each input sequence.

Caution

Local alignment does not attempt to align all bases between the two sequences. Instead, it compares segments of all possible lengths between the two sequences to optimise the similarity score that it calculates. Only the best-matching regions are aligned.

Note

Given a scoring scheme, the Smith-Waterman is guaranteed to find an optimal local alignment between two sequences.

Exercise 2 task

Using the same two DNA sequences as before, find their optimal local alignment using the EMBOSS Water program at EMBL-EBI.

2.3.1 Open the `EMBOSS Water` webpage

Open the link https://www.ebi.ac.uk/jdispatcher/psa/emboss_water in a web browser, to reach the landing page (Figure 2.8)

Figure 2.8: The `EMBOSS Water` landing page.

2.3.2 Enter the input sequences

Paste the two sequences you used for the Needle global alignment above into the two sequence input boxes, as shown in Figure 2.9.

Figure 2.9: The `EMBOSS Water` page with two input sequences provided.

2.3.3 Set appropriate parameters

The two sequences you have entered are DNA sequences, so check the radio button marked DNA (Figure 2.10)

Figure 2.10: `EMBOSS Water` input parameter set to `DNA`.

2.3.4 Run the alignment

Click on the Submit button at the bottom of the page to run the alignment. You can give your job a title, if you like (Figure 2.11).

Figure 2.11: `EMBOSS Water` submit button.

A holding window will appear while the job runs (Figure 2.12), and after a short wait you will be presented with the results page (Figure 2.13).

Figure 2.12: `EMBOSS Water` holding window.

Figure 2.13: `EMBOSS Water` results window.

2.3.5 Interpret the results (MyPlace Questions)

The key result returned by Water is again the overall sequence alignment, which once more is provided at the foot of the results page:

sequence_1        27 ACTTTAGAGGCTTGCAACCA     46
                     ||||||||||||||||||||
sequence_2         3 ACTTTAGAGGCTTGCAACCA     22

What do you notice about the alignment?

Although the input sequences are the same ones that were used for global alignment, the local alignment is quite different: it is much shorter.
Not all bases from either sequence are represented in the alignment - only bases 27 to 46 of sequence_1 and bases 3 to 22 of sequence_2.
Not all bases aligned in the global pairwise alignment are aligned in the local alignment.

Important

Please answer the questions below in the formative quiz on MyPlace

MyPlace formative quiz

Clicking on the green box should give you a hint to the answer, or where to find it.

What is the reported percentage sequence identity for the Needle global alignment?

Examine the second boxed section in the Needle output.

What is the reported percentage sequence identity for the Water local alignment?

Examine the second boxed section in the Water output.

Which alignment - global or local - do you think does the best job of aligning these sequences? Justify your answer with one or two sentences.

Consider the kinds of biological questions you might ask. When do you need to have the context of the full sequence information? When do you only need to know the best matching region? Which percentage identity result best represents the extent to which the two sequences actually resemble each other?

2.4 Summary

Well done!

After completing this section you should:

understand the difference between global and local pairwise alignments, and how to interpret their results
know how to use the EMBOSS Water and Needle tools to obtain optimal pairwise alignments
be able to explain the output from the Water and Needle tools

Now you are ready to move on to using BLAST, which uses local sequence alignments to search large databases of reference sequences to find the best matches to an input query sequence.