Protein Sequence Alignment Primary Structure Analysis Part 1

DNA takes us only so far • DNA sequencing relatively fast & cheap; but to get meaningful information from sequence, need: – to be able to distinguish genes from junk (a topic we’ll explore in some depth later on) – to be able to identify regulatory sequences – to be able to determine the function of a gene’s protein product – this will be our focus

Primary structure databases: nucleotides vs. proteins • For nucleotide sequences, GenBank is primary repository for data – source scientists have full authority over data – strict historical/archival point of view – companion database, RefSeq, is reviewed/corrected/annotated version of GenBank

• For protein sequences, there’s SwissProt (UniProt): an entirely different approach to database curation

The SwissProt Way

• SwissProt is not a primary repository like GenBank; instead, it is curated, primarily by a single person (Swiss scientist Amos Bairoch) – entries changed when new information available; flexible, correctable – considered best-annotated protein database – lot of work for one person, or even team of experts

• URL: http://www.expasy.ch/sprot/

TrEMBL • TrEMBL is SwissProt’s buffer database, similar in function (if not mechanics) to GenBank: GenBank is to RefSeq as TrEMBL is to SwissProt – consists of entries automatically derived from translation of DNA ORFs; data goes here before it goes to SwissProt – most of the protein sequences in both databases have never been isolated in nature; they are all derived from translated data

Searching SwissProt • Like GenBank, provides variety of access points • Starting point can be name of protein, or gene, or condition of interest • Can limit search by specific field (organism, protein name, e.g.)

Example: HER-2 positive breast cancer • We can start with a very general search:

• This produces almost 3000 hits; we can narrow this to our species:

Example continued • But this only cuts out about 1000 entries • Since we know (or think we know) the name of the protein, we can try adding another qualifier:

Example continued • This results in one relevant entry:

A closer look

Proteomics • Science of visualization & quantification of set of proteins present in given tissue or organism • Points of reference: – gel electrophoresis: separation of protein molecules by mass & charge – ORF translation: derive AA chain from nucleotide chain

• These often don’t match: more to protein structure (even primary structure) than simple translation tells us

Post-translational modification • Protein maturation process: modification(s) of primary structure that lead to ultimate tertiary/quaternary structure found in nature • Includes some combination of: – – – –

cuts within AA chain removal of AA fragments within chain chemical modifications of single AAs addition of lipid or sugar molecules

• Storage & retrieval of post-translational modification information is major role of protein databases

Location, location, location • Protein function related to its location – translation process involves exposing developing peptide chain to various chemical signals that specify location of mature protein – translocation: transport of protein across one or more membranes

Final destinations include: • attachment to cell membrane • secretion outside cell • transportation to mitochondria or other organelle • transportation to nucleus

Folding • Most important step in making mature protein – compacts peptide chain into stable 3D structure – final structure usually consists of several relatively independent domains;thousand of known domains • most proteins contain up to 10 • identifiable by scaffolded sequence signatures, or motifs, recognizably preserved over millions of years of evolution • domain architecture important because hints at 3D structure

Proteins vs. genes • Protein primary structures relatively simple compared to genes – AA sequences fairly short (average protein is 350 AAs) – have clear start & end – defined on single strand – although modifications can & do occur between ORF & mature protein, AA order remains stable

Using bioinformatics to determine protein function • CFTR protein serves as illustrative model: – Specific genetic defects can be identified in specific CF patients – Such genetic defects can be shown to lead to functional defects in CFTR – Next step: develop specific drugs to target specific defects (still in experimental stage)

• Problem: given DNA sequence of gene, how do we find cellular function of protein product?

Model organisms • Several organisms are known to have easily identifiable & mutable genes: • These organisms can serve as model organisms for investigation of gene behavior in humans where human genes have recognizable counterparts in the model

Structural clues to protein function • Proteins with similar primary structure (amino acid sequence) will likely have similar function • Similarity doesn’t have to extend to entire protein; can be more localized, e.g. regions of unknown protein sequence may resemble functional regions of known protein

Aligning protein sequences • Can be more effective for discovery than nucleotide sequence alignment • Sequence similarity often provides clues in function of unknown protein

Alignment scoring & evolution • Mutation is random process, but biological factors affect which mutations we actually see • We are most likely to observe the substitution of an amino acid with one that is chemically similar, because drastic change that disrupts protein function is likely to be selected against • Protein substitution matrix allows alignment algorithms to consider substitution likelihood to give better alignments

Protein similarity • Proteins much more complex than DNA; this works in our favor in terms of making effective comparisons • Amino acid similarity examples: – Aspartate and Glutamate have hydrophilic side chains – Leucine and Valine have hydrophobic side chains – A hydrophobic – hydrophilic substitution is more likely to alter protein function than phobic-phobic

Protein similarity • Can score not only exact matches but also conservative substitutions (mutations that result in functionally similar amino acids) • Such substitutions are more likely because they wouldn’t be selected against in evolution • Given all of the above, we can be confident of less ambiguity in protein alignment than in nucleotide alignment

Determining likelihood of substitution • Method 1: Look at chemical properties of amino acids: – hydrophobic vs. hydrophilic – charge – size of side chains

• Method 2: look at frequency of actual substitution occurrence in known sequences based on comparison of similar proteins – this is basis for substitution matrix • We will examine both methods

Method 1: Biochemical analysis of proteins • The Swiss Institute of Bioinformatics maintains ExPASy, a set of online tools for protein structure & function analysis • ExPASy stands for Expert Protein Analysis System • Two of the tools at ExPASy are ProtParam and ProtScale

ProtParam • Provides computation of physical & chemical properties of proteins from either user-entered raw sequence or from known entries in SwissProt/Trembl databases • Analysis includes: – number of amino acids in sequence – molecular weight – amino acid composition (% of total) – extinction coefficient (used for spectrophotometic analysis) – half-life: amount of time it takes for half of protein to degrade after synthesis – instability index

Primary structure analysis • Why analyze primary structure? – Need to take into account amino acid interactions to get clearer picture of secondary, tertiary structural factors – Segments with particular compositional types give clues to eventual conformation: • hydrophobic: potential transmembrane or core feature • coiled-coil: potential protein-protein interaction site • hydrophilic: potential surface structure

Primary structural analysis • Sliding window technique – Oldest sequence analysis method – Uses tables of amino acid properties: scale values

• Method – Pick window size based on desired feature: • for transmembrane feature: 19 • for globular feature: 7-11

– With window centered on one amino acid, scale values associated with all amino acids in window are summed & averaged, then result is associated with central AA – Shift window & continue until end of sequence – When finished, values associated with each AA are plotted against sequence: property profile

ProtScale

• An example of an online tool for performing sliding-window technique on a protein is ProtScale, also found in the ExPASy suite • The direct link to this resource is: http://www.expasy.org/cgi-bin/protscale.pl Note the extension – this is a perl program

Using ProtScale • Paste in FASTA sequence or type in accession number • Choose scale/window size • Click submit

Interpreting results • Consider only strong signals • Check signal robustness by repeating comparison using different scale

Sliding-window method: pros & cons • Advantage: relatively robust (not sensitive to scale changes) • Disadvantages: – not precise – window size is arbitrary

• Could go either way: does not interpret results for you

Testing for transmembrane segments in proteins • What transmembrane segments indicate: – One transmembrane segment at N-terminus of sequence suggests protein is secreted – Several transmembrane segments suggests a channel

• Can perform analysis with ProtParam, but more precise tool is TMHMM, which uses hidden Markov models (a sophisticated computational technique) to predict transmembrane regions

TMHMM

Link: http://www.cbs.dtu.dk/services/TMHMM-2.0/

TMHMM results • Predictions are precise • Predicts segments inside/outside cell

Looking for coiled-coil segments • Coiled-coils are regions formed by intertwining alpha-helices • May indicate protein-protein interaction site • May also lead to false results in database searches, so it’s good to know location in case you need to filter out • Online tool available at: http://www.ch.embnet.org/software/COILS_form.html