Protein Sequence Analysis for bioinformatics

Proteins are the workers of the cell. They carry oxygen, fight infections, copy DNA, and send messages from one cell to another. Every protein begins as a chain of amino acids, and that chain is represented by a sequence of letters, such as MALWMRLLPL.... This is called a protein sequence. A protein sequence can tell us a lot about a protein. Using computer tools, we can estimate its molecular weight, electric charge, signal peptide, transmembrane regions, and possible secondary structure. These predictions help researchers study proteins without immediately performing laboratory experiments.

In this article, we will explore these concepts step by step using human insulin as an example. You can follow the steps on your computer as you read.

Part 1: Getting Started with Sequences and ExPASy Tools

What Is a Protein Sequence?

There are 20 common amino acids, and each has a one-letter code. For example, A represents alanine, L represents leucine, K represents lysine, and E represents glutamic acid. A protein sequence is a series of these letters arranged from the beginning (N-terminus) to the end (C-terminus) of the chain. Protein sequences are commonly stored in a text format called FASTA. It contains a header line beginning with the > symbol, followed by the sequence.

>sp|P01308|INS_HUMAN Insulin OS=Homo sapiens OX=9606 GN=INS PE=1 SV=1 MALWMRLLPLLALLALWGPDPAAAFVNQHLCGSHLVEALYLVCGERGFFYTPKTRREAEDLQVGQVELGGGPGAGSLQPLALEGSLQKRGIVEQCCTSICSLYQLENYCN

This is the sequence of human preproinsulin, which contains 110 amino acids. We will use it as our main example throughout the article.

Where Can We Find Protein Sequences?

The UniProt database provides curated protein information and sequences. Each protein entry has a unique identifier called an accession number. Human insulin has the accession number P01308.

To retrieve the sequence:

1. Open the UniProt website.
2. Search for P01308.
3. Open the human insulin entry and select the FASTA option.
4. Copy the sequence for use in the tools discussed below.

What Is ExPASy?

ExPASy is a bioinformatics resource portal maintained by the SIB Swiss Institute of Bioinformatics. It provides tools that help researchers analyze proteins and other biological data.

The website interface may change over time. If a tool is difficult to locate, search for its name on the ExPASy portal.

Part 2: Physicochemical Properties, pI, and Molecular Weight

Physicochemical properties describe the physical and chemical characteristics of a protein. Two important properties are molecular weight and isoelectric point.

Molecular Weight (MW)

Molecular weight indicates the mass of a protein. It is commonly expressed in daltons (Da) or kilodaltons (kDa), where 1 kDa equals 1,000 Da.

As a rough estimate, an average amino acid residue contributes about 110 Da to a protein. Therefore, a protein containing 100 amino acids may have a molecular weight of approximately 11 kDa.

Why is molecular weight important?

• It helps researchers interpret protein bands in SDS-PAGE gels.
• It supports the selection of suitable protein purification methods.
• It helps compare predicted protein masses with experimental measurements from mass spectrometry.

Isoelectric Point (pI)

Proteins contain amino acids that can carry positive or negative charges. Lysine, arginine, and histidine can carry positive charges, while aspartic acid and glutamic acid can carry negative charges.

The isoelectric point (pI) is the pH at which a protein has no net electrical charge.

• Below its pI, a protein generally has a net positive charge.
• Above its pI, a protein generally has a net negative charge.
• At its pI, the positive and negative charges balance overall.

The pI can help researchers select buffers, understand protein solubility, and plan ion-exchange chromatography or two-dimensional gel electrophoresis.

Worked Example: Analyzing Insulin with ProtParam

1. Open the ExPASy ProtParam tool.
2. Paste the human preproinsulin sequence without the FASTA header.
3. Select the option to calculate the parameters.
4. Review the results.

Approximate results for the 110-amino-acid preproinsulin sequence include:

These values are approximate and should be checked against the tool's current output. The theoretical pI is below 7, which is consistent with the sequence having more acidic residues (D and E) than basic residues (K and R).

Mature insulin is smaller than preproinsulin. After processing, it consists of two chains: the A chain, containing 21 amino acids, and the B chain, containing 30 amino acids. Together, the mature hormone has a molecular mass of approximately 5.8 kDa.

The two chains are connected by disulfide bonds. When calculating the masses of the separate chains, the total may differ slightly from the mass of mature insulin because bond formation involves the loss of hydrogen atoms.

A chain:

GIVEQCCTSICSLYQLENYCN

B chain:

FVNQHLCGSHLVEALYLVCGERGFFYTPKT

This example shows why it is important to distinguish a protein's initial precursor from its mature, biologically active form.

Other Useful ProtParam Results

Practice: Analyze the insulin A and B chains separately using ProtParam. Compare their predicted pI values and consider how their amino acid compositions influence the results.

Part 3: Signal Peptide Prediction

What Is a Signal Peptide?

A signal peptide acts like an address label for a protein. It is usually a short sequence near the beginning of a newly made protein that directs it toward a particular cellular destination, often the secretory pathway or a membrane.

Proteins entering the secretory pathway pass through the endoplasmic reticulum (ER) and often the Golgi apparatus. These include many hormones, antibodies, digestive enzymes, and membrane proteins.

After a protein enters the ER, an enzyme called signal peptidase may remove its signal peptide. The remaining sequence can then undergo further processing to form the mature protein.

What Does a Signal Peptide Look Like?

Many signal peptides contain three general regions:

1. N-region: A short region that may contain positively charged amino acids such as lysine (K) or arginine (R).
2. H-region: A hydrophobic region containing amino acids that interact well with oily environments.
3. C-region: A region near the cleavage site that helps determine where the signal peptide is cut.

Insulin provides an example of a protein synthesized with a signal peptide. In human preproinsulin, the signal peptide occupies the first 24 amino acids. After it is removed, the remaining proinsulin undergoes additional processing before mature insulin is formed.

Predicting Signal Peptides with SignalP

SignalP is a tool used to predict signal peptides and their potential cleavage sites from protein sequences.

To try it:

1. Open the SignalP web server.
2. Paste the protein sequence.
3. Select the appropriate organism group, such as eukaryotes for human proteins.
4. Submit the sequence and review the prediction.

For human preproinsulin, you would expect the tool to identify a signal peptide near the beginning of the sequence, with a cleavage site around position 24. Exact results depend on the tool version and model.

How to Interpret SignalP Results

• Prediction score: Indicates how strongly the sequence matches the model's signal peptide prediction.
• Cleavage site: The predicted position where the signal peptide is removed.
• Organism group: The selected group matters because signal peptides can differ between organisms.

Tools such as Phobius can also predict signal peptides and transmembrane regions, helping distinguish between similar-looking hydrophobic sequences.

Practice: Analyze a protein that is normally found inside the cell, such as human GAPDH (UniProt accession P04406). Compare its result with the insulin prediction.

Part 4: Transmembrane Prediction

What Are Transmembrane Proteins?

Cell membranes contain a lipid bilayer with a hydrophobic interior. A transmembrane protein must have regions that can interact with this environment. Many transmembrane regions form alpha helices containing approximately 18–25 mostly hydrophobic amino acids.

Some proteins cross the membrane once, while others cross it multiple times. Many receptors involved in detecting hormones, light, and other signals belong to the second group.

Transmembrane predictions help researchers understand protein structure and function, including the possible locations of drug-binding sites.

Method 1: Hydropathy Plot with ProtScale

A hydropathy plot shows how hydrophobic or hydrophilic different parts of a protein sequence are. One commonly used method is the Kyte–Doolittle scale.

The method assigns a hydropathy score to each amino acid and calculates average scores across a sliding window. A strong hydrophobic region may appear as a peak in the plot.

To try it:

1. Open ExPASy ProtScale.
2. Paste your protein sequence.
3. Select the Kyte–Doolittle hydropathy scale.
4. Choose a window size of about 19 residues and generate the plot.

For preproinsulin, a hydrophobic region appears near the beginning because its signal peptide contains many hydrophobic amino acids. However, this peak does not indicate a true transmembrane helix. It demonstrates why hydropathy plots should be interpreted carefully.

Method 2: Modern Transmembrane Predictors

Tools such as DeepTMHMM and TMHMM predict transmembrane regions. Phobius can also help distinguish signal peptides from transmembrane helices.

A useful example is the human beta-2 adrenergic receptor (UniProt accession P07550), which is a seven-transmembrane receptor.

To analyze it:

1. Retrieve its FASTA sequence from UniProt.
2. Submit the sequence to DeepTMHMM or another available predictor.
3. Examine the predicted membrane-spanning regions and topology.

The receptor is expected to have seven transmembrane helices. The precise prediction should be checked against the tool's output.

How to Read the Results

A Note About Hydrophobic Regions

A hydrophobic sequence may be a signal peptide, a transmembrane helix, or a region buried inside a folded protein. It is best to compare the results of multiple tools and consider the protein's known biological role.

Practice: Compare the hydropathy plots of preproinsulin and the beta-2 adrenergic receptor. Notice how their hydrophobic regions differ.

Part 5: Secondary Structure Prediction

What Is Secondary Structure?

Protein secondary structure describes local folding patterns in a protein chain. The three common categories are alpha helices, beta strands, and coils or loops.

These local structures contribute to the overall three-dimensional shape of a protein.

How Do Computers Predict Secondary Structure?

Amino acids differ in their tendency to form particular structures. For example, alanine and leucine often occur in helices, while valine and isoleucine frequently occur in beta strands. Proline can disrupt helices, and glycine's flexibility can support turns and loops.

Several methods are used for secondary structure prediction:

1. Chou–Fasman method: Uses amino acid preferences to estimate structural tendencies.
2. GOR method: Considers an amino acid and its neighboring residues.
3. Modern machine learning methods: Tools such as PSIPRED and JPred use sequence patterns and evolutionary information to predict structural regions.

PSIPRED and JPred are useful starting points for beginners. Prediction accuracy varies depending on the protein, available sequence information, and the method used.

Worked Example: Predicting Insulin's Secondary Structure

1. Open PSIPRED or JPred.
2. Paste the protein sequence. You can analyze the full preproinsulin sequence or examine the A and B chains separately.
3. Submit the sequence and wait for the result.
4. Review the predicted structural states and confidence scores.

A result may represent each amino acid with a letter such as H, E, or C. For example, the following is only an illustration, not an actual prediction:

Sequence:    GIVEQCCTSICSLYQLENYCN Prediction:  CHHHHHHHCCHHHHHHHHCCC

Some tools also display diagrams using cylinders for helices, arrows for strands, and lines for coils.

How to Evaluate a Prediction

• Review confidence scores where available.
• Compare predictions from different tools.
• Compare the result with an experimentally determined structure in the Protein Data Bank (PDB), when available.
• Remember that predictions are estimates and may differ from experimental observations.

Experimental methods such as X-ray crystallography, nuclear magnetic resonance (NMR), and cryo-electron microscopy can provide structural evidence.

Practice: Predict the secondary structure of the insulin B chain using two tools. Record the regions where the predictions agree and differ.

Putting It All Together: A Mini Project

You can combine these tools to build a basic profile of human insulin.

These analyses help explain how a protein's sequence relates to its biological role. Preproinsulin contains a signal peptide and is processed into proinsulin and then mature insulin. Mature insulin consists of two chains linked by disulfide bonds.

Choose another protein that interests you and repeat these steps. Write a short report describing the sequence, predicted properties, and possible biological function.

Conclusion

Protein sequence analysis is an important starting point in proteomics and bioinformatics. Tools such as UniProt, ExPASy ProtParam, SignalP, DeepTMHMM, PSIPRED, and JPred help researchers examine protein properties, predict cellular destinations, identify membrane-spanning regions, and estimate secondary structures.

By practicing with these tools, beginners can learn how to interpret protein sequences and connect computational predictions with biological functions. Always record the tools and versions used, compare predictions where possible, and remember that computational results are not a substitute for experimental confirmation.