Pharm 207/Bio 207  Using Internet Resources in Molecular Biology - Lecture 2  DNA & Protein Sequence Comparison

Introduction

DNA and protein sequences are being determined at an ever increasing rate. These data are fundamental to bioinformatics as we know it today. The area of bioinformatics known as sequence analysis concerns the discovery of biological knowledge from these sequences. Typically the user begins with one or more sequences (the probe(s)) for which they wish to learn something about the biology of the organism from which the sequence(s) was obtained.. This involves (at least as far as this lecture is concerned) comparison to itself and one or more other sequences from which comparative information is inferred. For this inference to be correct the detailed annotation of the known sequences must be correct and the comparison must be deemed to exist. This become more problematic as more of the annotation is itself derived from comparison rather than direct biochemical evidence - a natural consequence resulting from the amount of sequence data being generated..  

The Basics

• Comparison of an unknown sequence to an annotated sequence permits us to infer structural, functional and evolutionary relationships.
• Wherever possible use the protein sequence since this confers more information.
• Sequence alignment provides explicit mapping between sequences - pairwise implies 2 sequences only
• This comparative process become more valuable the more sequences there are that have been annotated.
• As the size of the databases grows more efficient algorithms are needed.
• Note the distinction between similarity and homology:
  • Similarity is simply a measure of expression how alike two sequences are
  • Homology means there is an evolutionary relationship between two sequences. There are not degrees of homology.
  • Extending this to individual residues they are either identical or similar - similar implies that they at least share certain properties.
• Substitutions, deletions and insertions all occur as part of the natural evolutionary process.

The search for homology is well supported by internet accessible tools, but getting the correct answer with these tools requires that one understand something about the relationship between sequence alignments, database searches, and the statistics of sequence matching.  Related methods for finding sequential or structural motifs require similar understanding.

The Basics - Amino Acid Properties

To understand substitution matrices and other features of protein sequence pairwise alignment it is necessary to understand the basic properties of amino acids.

The Single-Letter Amino Acid Code

Amino Acid Structures

All amino acids have the same general formula:

The twenty amino acids found in biological systems are:

Courtesy Nancy Watson at nwatson@ilstu.edu

More details on amino acids can be found in the Principles of Protein Structure Using the Internet.

It is possible to displays values for static properties on a percentage scale as defined by R.A. Bogardt, et al. (J. Mol. Evol. 15, 197-218, 1980).

The values used are as follows:

Amino Acid    Volume    Polarity     Isoelectric Point    Hydrophobicity     Mean Sol. Accessibility

Gln                   51.3         45.7                 34.4                         0.0                         43.6
Ser                   18.1         32.1                 34.8                         1.9                         40.5
Thr                   34.0         21.0                 45.7                         1.9                         35.3
Asn                  35.4         63.0                 31.3                         2.4                         46.1
Gly                     0.0         37.0                 38.5                         2.7                         54.0
Asp                  31.3       100.0                 0.0                         17.5                         45.0
Glu                   47.2         93.8                 3.2                         17.8                         48.6
Arg                   70.8         51.9             100.0                         22.6                         50.1
Ala                    15.9         25.9               39.2                         23.1                         37.4
His                     49.2        43.2                59.2                         23.1                         28.1
Cys                    28.0         7.4                 26.3                         40.3                         7.4
Lys                    68.0        64.2                 86.9                         43.5                         54.3
Val                     47.7         8.6                 38.5                         49.6                         19.6
Met                     62.8         4.9                 35.7                         44.3                         3.9
Leu                     63.6         0.0                 38.6                         57.6                         10.1
Tyr                     78.5         9.9                 34.4                         70.8                         30.1
Pro                     41.0         21.0             40.2                         73.5                         66.2
Phe                     77.2         1.2             38.6                             76.1                         5.5    
Ile                     63.6         0.0                 39.2                         83.6                             7.5
Trp                     100.0     4.9                 37.7                         100.0                         13.8


Yet more information from the Jena Image Library of Biological Macromolecules

Practical: Manual Sequence Alignments

Align the following sequences by hand using what you know about amino acid properties.

Use:

|   for a perfect match

.   for a conservative substitution

    (blank) for no match

- for a gap in one or other of the sequences.

1.  Align ANALYSIS and ANALYZED (no gaps)

2. Align ANALYSISPRIMER and ANALYZEDERIVED (no gaps)

3. Align VVVVVASCDEFGYYYYY and  QQQQQATCEEFGLLLLL

(Think about this in the context of a local versus global (total) alignment

4. Align TKTYFPHFDLSHGSAQVKGHGKKV and TQRFFESFGDLSTPDAVMGNPKVKAHGKKV

(gapped)

More Detailed Concepts [See BIMM140 for further details]

Homology versus Similarity

• homology:
  • the presence of a similar feature because of descent from a common ancestor
• homoplasy - the presence of a similar feature because of convergence
• similarity - quantitative measure of 'likeness' of two or more sequences
• similarity - use statistics to determine 'significance' of the similarity

Homology

• Homology cannot be observed.
  • We can’t actually see the ancestral organisms/molecules and trace descent.
  • These ancestral organisms / molecules no longer exist in today's world.
• Homology is an inference, a conclusion we draw based on observed similarity.
  • In practice, if the similarity statistics indicate a high enough degree of significance, one infers that the two sequences are homologues
  • As homologues, one implicitly infers common ancestry for the two sequences.
  • It is, therefore, subject to degrees of certainty.
• Homology is an all-or-none relationship
  • Homology is like Pregnancy: one is ... or one isn't ... all-or-none
• Homology strongly suggests similar structure and function for the proteins
• Significantly similar molecular sequences are very unlikely to arise by chance - i.e. homoplasy on the molecular level is very unlikely
  • Caveat: horizontal transfer of sequences from one organism to another
• Because molecular homoplasy is unlikely, significant sequence similarity is a strong indication of homology

 Global and Local Alignment

• An alignment that spans the whole length of the overlapping sequences is considered a global alignment.
• This may not be optimal given that proteins exist as domains and given sequences may contain multiple domains, often rearranged. As we shall see the dotplot provides a mechanism for revealing multiple regions of local alignment.
• A global alignment is useful when you know you have a single domain and sequences have not diverged substantially.

Alignments in General - Dynamic Programming

• There are many alignments thus it is necessary to define a score so that optimal alignments can be recognized based on that scoring scheme. The most optimal alignment is the one with the highest score.
• The score is usually derived by summing incremental contributions at each step along the way.
• One method for the determination of the best path strategy is known as dynamic programming and has its roots in graph theory.
• This was first applied by Needleman and Wunsch and published in 1970 in 1970 in perhaps the most cited paper in bioinformatics.
• The idea is one of extension of optimal subpaths to provide a complete (global) path.
• An extension of the original work to provide local alignments simply says the alignment need not reach the edges of the search graph. Rather  a local alignment is extended and stops if the score becomes zero, at which point a new local alignment is started. This may lead to a number of local alignments, the one with the best score is reported as the optimal local alignment.
• The optimal local alignment may not be the most biologically meaningful - it pays to review a number of sub-optimal alignments.

Scores and Gap Penalties

• The scoring function is critical to determining good alignments.
• The absolute match/mismatch rule for scoring works okay for DNA sequences but for proteins does not take advantage of the observation that some amino acid substitutions are more likely than others - these are referred to as conservative substitutions.
• These substitutions conform to the property groups we have discussed already - examples isoleucine for valine (both small and hydrophobic) and serine for threonine (both polar).
• Scoring is such as to provide a scale of scoring for absolute conservation, versus conservative substitution versus non-conserved substitution. This is manifest in a substitution matrix. [More on matrices from BIMM 140]
• Gaps represent insertions and deletions in one or other of the sequences being aligned.
• Too large a gap may represent an implausible set of mutations - in other works could imply a loss of biological function.
• Gap penalties are defined by two parameters - the gap opening penalty and the gap extension penalty The rationale here (which is empirical) is that it is difficult to open a gap, but when it is open it is not so difficult to extend it.

Statistical Significance of Alignments

• It is easy to define a score, but how do you know that score represents something biologically meaningful ie some evidence of homology.
• Such a statistical treatment is difficult for global alignments, but better for local alignments.
• Statistics derived from local alignments without gaps - segments referred to as a high-scoring segment pair (HSP).
• The probability density function used to describe the behavior of HSPs is known as the extreme value distribution which differs from a normal value distribution.
• By relating a score S to the expected distribution it is possible to calculate a p value which indicates the probability that the specified score could occur by chance.
• Thus p values close to zero are the most significant.
• A related quantity is E - the expectation value - the expected number of chance alignments that would achieve a score of S or higher. The smaller the value of E the more statistically significant the alignment.

Database Searching

• Pairwise sequence alignments can be extended such that one sequence becomes the query sequence and any of the sequences in a database become a target sequence to determine a large set of pairwise alignments.
• To scale computationally either parallel computation is needed or heuristic methods.
• Heuristic methods make assumptions at the risk of missing some alignments.
• Word based approaches are the most popular - rather than comparing individual amino acids or nucleotides words (groupings of amino acids or nucleotides) are compared.
• Both FASTA and BLAST - two programs used in this workshop use the concept of words and will be discussed in more detail.

Filtering

• Both protein and DNA sequences contain low-complexity regions and DNA sequences in particular contain repetitive elements.
• Filtering attempts to remove the confusion such regions can cause to a database search.
• This can be done either by first checking the query sequence or filtering the results.

 Dot Plots

Dot plots are probably the oldest way of comparing sequences (Maizel and Lenk). A dot plot is a visual representation of the similarities between two sequences. Each axis of a rectangular array represents one of the two sequences to be compared. A window length is fixed, together with a criterion when two sequence windows are deemed to be similar. Whenever one one window in one sequence resembles another a window in the other sequence, a dot or short diagonal is drawn at the corresponding position of the array. Thus, when two sequences share similarity over their entire length a diagonal line will extend from one corner of the dot plot to the diagonally opposite corner. If two sequences only share patches of similarity this will be revealed by diagonal stretches.

Figure 1 shows an example of a dot plot. There, the alpha chain of human hemoglobin is compared to the beta chain of human hemoglobin. For this computation, the window length was set to 31, matches and mismatches were assigned similarity values of +5 and -4 respectively. The gray values of the dots scale with the similarity of two windows. One can clearly discern a diagonal trace along the entire length of the two sequences. Note the jumps where this trace jumps to another diagonal of the array. These jumps correspond to position where one or the other sequence has more (or less) letters than the other one.

Alignment of the A and B protein chains of deoxy hemoglobin.

 
Dot plots are a very powerful method of comparing two sequences. They do not predispose the analysis in any way such that they constitute the ideal first-pass analysis method. Based on the dot plot the user can decide whether he deals with a case of global, i.e. beginning-to-end similarity, or local similarity. Local similarity denotes the existence of similar regions between two sequences that are embedded in the overall sequences which lack similarity. Sequences may contain regions of self-similarity which are frequently termed internal repeats. A dot plot comparison of the sequence will itself will reveal internal repeats by displaying several parallel diagonals. 

Exercise 1

1. Download and install windot (read the dotplot text file to see what to do)

2. Download the sequences for the A and B chains of hemogblobin

• Go to the PDB
• Enter 4HHB
• Go to sequence details
• Download the A and B chains at FASTA files
• Experiment with different window sizes and thresholds

Go to NBCI and get your pet protein sequence and compare it to itself in the same way. Describe your findings.

{NOTE RED represents the minimum of what I expect to see described in your final documentary}

BLAST

 Tutorial 1

 Tutorial 2

PSI-BLAST

Tutorial

Exercise 2

For your pet protein sequence run against psi-blast using different scoring matrices. Attempt to optimize the number of remote homologs found and report on the biological implications of your findings. Include details of any interesting alignments in your report.

Reading Materials

• Required Chapter 7 of A.D. Baxevanis and B.F. Ouellette. "Bioinformatics: A Practical Guide to the Analysis of Genes and Proteins". John Wiley, 1998.
• Optional Chapter 6 of Attwood and Parry-Smith "Introduction to Bioinformatics". Addison Wesley Longman (UK), 1999. ISBN: 0 582 327881

1.   Molecular Biology Workbench - General Purpose Sequence Analysis Tool
2. Dotter program
3. Database Searches
   • NCBI BLAST
   • EBI FASTA
   • PIR BLAST and FASTA
4. General Information
   Lists of many sites and tools
   • DNASYSTEM
   • CMS MBR

   ---