Here are examples of algorithms commonly used in bioinformatics for multiple sequence alignment (MSA) and related tasks:

1. MAFFT (Multiple Alignment using Fast Fourier Transform): MAFFT is a popular algorithm for MSA that uses an iterative refinement approach. It is known for its speed and accuracy, especially for aligning large datasets.
2. Clustal Omega: Clustal Omega is another widely used MSA algorithm that offers improved speed and scalability compared to earlier versions like ClustalW. It uses a progressive alignment approach with heuristics to improve alignment quality.
3. MUSCLE (Multiple Sequence Comparison by Log-Expectation): MUSCLE is known for its speed and accuracy in MSA. It uses a progressive alignment approach followed by iterative refinement to improve the alignment quality.
4. T-Coffee (Tree-based Consistency Objective Function for Alignment Evaluation): T-Coffee is a versatile MSA algorithm that uses a combination of pairwise and multiple sequence alignments to generate high-quality alignments. It is known for its ability to handle complex alignment problems.
5. PSI-BLAST (Position-Specific Iterative Basic Local Alignment Search Tool): PSI-BLAST is a variant of the popular BLAST algorithm that iteratively builds a position-specific scoring matrix (PSSM) based on the initial alignment. It is used for searching protein sequence databases and for detecting remote homologs.
6. HMMER: HMMER is a suite of algorithms for protein sequence analysis that includes tools for profile hidden Markov model (HMM) creation and searching. It is commonly used for protein domain identification and classification.
7. ProbCons: ProbCons is an algorithm that uses a probabilistic consistency-based approach to align sequences. It iteratively refines the alignment based on a probabilistic model, which can improve alignment accuracy, especially for distantly related sequences.

These algorithms are widely used in bioinformatics for tasks such as phylogenetic analysis, protein structure prediction, and functional annotation. Each algorithm has its strengths and weaknesses, and the choice of algorithm depends on the specific requirements of the analysis.

Consistency-based Methods

Explanation of consistency-based approaches

Consistency-based approaches are used in multiple sequence alignment (MSA) to improve the accuracy of alignments, especially for distantly related sequences. These approaches are based on the principle that if two sequences are homologous (i.e., share a common ancestor), they should have similar patterns of conservation and variation across their lengths. Here’s an overview of how consistency-based approaches work:

1. Pairwise Alignments: The process starts with pairwise alignments between all pairs of sequences in the dataset. This step is typically done using dynamic programming algorithms such as Needleman-Wunsch or Smith-Waterman.
2. Consistency Scoring: For each column in the alignment, a score is computed based on the consistency of residue pairs. Conserved residue pairs (i.e., pairs of residues that are aligned in multiple pairwise alignments) receive a higher score, while variable residue pairs receive a lower score.
3. Consistency Transformation: The consistency scores are transformed into a similarity matrix, where high scores indicate strong consistency (i.e., residues that are likely to be homologous) and low scores indicate weak consistency.
4. Guide Tree Construction: Based on the similarity matrix, a guide tree (or guide alignment tree) is constructed using hierarchical clustering methods such as neighbor-joining or UPGMA. The guide tree represents the evolutionary relationships between sequences and serves as a guide for the progressive alignment process.
5. Progressive Alignment: The sequences are aligned progressively based on the guide tree, starting with the most closely related sequences and gradually adding more distant sequences to the alignment. During this process, the consistency scores are used to guide the alignment, ensuring that residues that are consistent across multiple pairwise alignments are aligned together.
6. Iterative Refinement: After the progressive alignment is completed, the alignment can be refined iteratively using methods such as iterative refinement or profile-profile alignment. This step further improves the alignment accuracy by iteratively realigning sequences based on the initial alignment.

Consistency-based approaches are particularly useful for aligning distantly related sequences, where traditional alignment methods may produce less accurate results. By incorporating information from multiple alignments, consistency-based approaches can improve the accuracy of alignments and provide better insights into the evolutionary relationships between sequences.

Examples of algorithms (e.g., T-Coffee)

Here are examples of algorithms that use consistency-based approaches for multiple sequence alignment (MSA):

1. T-Coffee (Tree-based Consistency Objective Function for Alignment Evaluation): T-Coffee is a popular MSA algorithm that uses a consistency-based approach to align sequences. It combines information from multiple sequence alignments, including pairwise alignments and structural information, to generate high-quality alignments. T-Coffee is known for its ability to handle complex alignment problems and produce accurate alignments, especially for distantly related sequences.
2. ProbCons: ProbCons is another MSA algorithm that uses a probabilistic consistency-based approach to align sequences. It iteratively refines the alignment based on a probabilistic model, which can improve alignment accuracy, especially for distantly related sequences. ProbCons is known for its accuracy and ability to handle large datasets.
3. MAFFT (Multiple Alignment using Fast Fourier Transform): While MAFFT is primarily known for its speed and scalability, it also incorporates a consistency-based approach in its iterative refinement process. MAFFT aligns sequences progressively and then iteratively refines the alignment based on consistency scores, improving alignment quality.
4. MUSCLE (Multiple Sequence Comparison by Log-Expectation): MUSCLE uses a consistency-based approach in its iterative refinement process to improve alignment accuracy. It starts with a progressive alignment and then iteratively refines the alignment based on consistency scores, producing more accurate alignments, especially for distantly related sequences.
5. PCMA (Probabilistic Consistency-based Multiple Alignment): PCMA is a MSA algorithm that uses a probabilistic consistency-based approach to align sequences. It combines information from multiple alignments to generate a consensus alignment, improving alignment accuracy and robustness.

These algorithms are widely used in bioinformatics for tasks such as phylogenetic analysis, protein structure prediction, and functional annotation. They leverage consistency-based approaches to improve alignment accuracy, especially for sequences that are evolutionarily distant.

Statistical Alignment

Introduction to statistical methods

Statistical methods play a crucial role in bioinformatics and computational biology, providing tools for data analysis, hypothesis testing, and model building. These methods help researchers make sense of biological data, identify patterns, and draw meaningful conclusions. Here’s an introduction to some key statistical methods used in these fields:

1. Descriptive Statistics: Descriptive statistics are used to summarize and describe the features of a dataset. Measures such as mean, median, mode, variance, and standard deviation are used to describe the central tendency, dispersion, and shape of the data.
2. Inferential Statistics: Inferential statistics are used to make inferences or predictions about a population based on a sample of data. These methods include hypothesis testing, confidence intervals, and regression analysis. Hypothesis testing is used to determine whether an observed effect is statistically significant or due to random chance.
3. Statistical Hypothesis Testing: Statistical hypothesis testing is used to test the validity of a hypothesis by comparing observed data to expected values under a null hypothesis. Common tests include t-tests, chi-square tests, and ANOVA (analysis of variance).
4. Regression Analysis: Regression analysis is used to model the relationship between a dependent variable and one or more independent variables. It is often used to predict the value of the dependent variable based on the values of the independent variables.
5. Cluster Analysis: Cluster analysis is used to group similar data points into clusters based on their characteristics. It is often used in bioinformatics to identify patterns in gene expression data or to classify protein sequences based on their features.
6. Machine Learning: Machine learning techniques, such as supervised learning, unsupervised learning, and deep learning, are increasingly used in bioinformatics for tasks such as classification, regression, clustering, and feature selection. These methods can handle large and complex datasets and can extract patterns that may not be apparent using traditional statistical methods.
7. Bayesian Statistics: Bayesian statistics is a framework for statistical inference in which probabilities are assigned to hypotheses. It is particularly useful for integrating prior knowledge with observed data and for making predictions based on uncertain information.

These statistical methods are essential tools in bioinformatics and computational biology, enabling researchers to analyze complex biological data, test hypotheses, and uncover new insights into biological systems.

Examples of algorithms (e.g., ProbCons)

Here are examples of statistical algorithms commonly used in bioinformatics and computational biology:

1. ProbCons: ProbCons is a probabilistic consistency-based algorithm for multiple sequence alignment. It uses a probabilistic model to guide the alignment process, incorporating information from multiple sequence alignments to improve accuracy.
2. Hidden Markov Models (HMMs): HMMs are statistical models used for representing sequence data, such as protein or DNA sequences. They are used in various bioinformatics applications, including sequence alignment, gene prediction, and motif finding.
3. Bayesian Networks: Bayesian networks are probabilistic graphical models that represent probabilistic relationships among variables. They are used in bioinformatics for gene regulatory network modeling, protein structure prediction, and other applications.
4. Phylogenetic Tree Construction Algorithms: Algorithms such as Neighbor-Joining, Maximum Likelihood, and Bayesian inference are used to construct phylogenetic trees from sequence data. These algorithms use statistical models to infer the evolutionary relationships between species or genes.
5. Statistical Learning Algorithms: Machine learning algorithms such as Support Vector Machines (SVM), Random Forests, and Neural Networks are used in bioinformatics for tasks such as classification, regression, and clustering of biological data.
6. Statistical Tests for Differential Expression Analysis: Statistical tests such as t-tests, ANOVA, and Fisher’s exact test are used to identify genes or proteins that are differentially expressed between experimental conditions in gene expression studies.
7. Markov Chain Monte Carlo (MCMC) Methods: MCMC methods are used for Bayesian inference and sampling from complex probabilistic models. They are used in bioinformatics for parameter estimation in phylogenetic analysis, protein structure prediction, and other applications.

These are just a few examples of the many statistical algorithms used in bioinformatics and computational biology. Each algorithm has its strengths and is used in specific contexts to analyze and interpret biological data.

Comparison of Different Algorithms

Performance metrics (e.g., accuracy, speed, scalability)

Performance metrics are crucial for evaluating the effectiveness and efficiency of algorithms in bioinformatics and computational biology. Here are some common performance metrics used to assess algorithms:

1. Accuracy: Accuracy measures how well an algorithm performs compared to a known or expected outcome. In bioinformatics, accuracy can be measured by comparing the algorithm’s results to a gold standard or by assessing its ability to correctly predict biological phenomena.
2. Sensitivity and Specificity: Sensitivity measures the proportion of true positives that are correctly identified by the algorithm, while specificity measures the proportion of true negatives that are correctly identified. These metrics are commonly used in binary classification tasks, such as identifying disease-associated genes.
3. Precision and Recall: Precision measures the proportion of true positive predictions among all positive predictions made by the algorithm, while recall measures the proportion of true positives that are correctly identified by the algorithm among all actual positives. These metrics are important for evaluating the performance of algorithms in tasks where false positives or false negatives are costly, such as in detecting disease-related mutations.
4. F1 Score: The F1 score is the harmonic mean of precision and recall and provides a single metric that balances both metrics. It is useful for evaluating algorithms that need to balance between precision and recall.
5. Speed: Speed refers to how quickly an algorithm can process data and produce results. In bioinformatics, where large datasets are common, speed is often a critical factor in algorithm performance.
6. Scalability: Scalability measures how well an algorithm can handle increasing amounts of data or complexity. Algorithms that can scale efficiently are important for analyzing large-scale biological datasets.
7. Robustness: Robustness refers to an algorithm’s ability to perform well under various conditions, such as noise or missing data. Robust algorithms are essential for handling real-world biological data, which can be noisy and incomplete.
8. Resource Usage: Resource usage metrics, such as memory consumption and computational resources required, are important for evaluating the practicality of an algorithm in terms of hardware requirements and cost.

These performance metrics are used to assess the quality and efficiency of algorithms in bioinformatics and computational biology, helping researchers choose the most suitable algorithms for their specific tasks.

Considerations for selecting an algorithm for specific applications

When selecting an algorithm for a specific application in bioinformatics or computational biology, several key considerations should be taken into account to ensure that the chosen algorithm is suitable for the task at hand. Some of these considerations include:

1. Accuracy: The algorithm should be capable of producing accurate results for the specific type of data and analysis being performed. Consider the algorithm’s performance metrics, such as sensitivity, specificity, precision, recall, and F1 score, to assess its accuracy.
2. Speed: For time-sensitive analyses or large-scale datasets, the speed of the algorithm is crucial. Consider the algorithm’s computational complexity and runtime efficiency to ensure that it can handle the size and complexity of the dataset within acceptable time limits.
3. Scalability: The algorithm should be able to scale efficiently with increasing dataset sizes or complexity. Consider the algorithm’s scalability in terms of memory usage, computational resources, and runtime performance as the dataset grows.
4. Robustness: The algorithm should be robust to noise, missing data, and other sources of variability commonly found in biological datasets. Consider the algorithm’s ability to handle these challenges and produce reliable results under real-world conditions.
5. Interpretability: The algorithm should produce results that are interpretable and meaningful to biologists and researchers. Consider how easily the algorithm’s output can be understood and integrated into biological knowledge.
6. Availability: The algorithm should be available in a form that is accessible and easy to use. Consider whether the algorithm is open-source, well-documented, and supported by a community of users and developers.
7. Compatibility: The algorithm should be compatible with the data format and analysis pipeline used in the specific application. Consider whether the algorithm can be easily integrated into existing workflows and tools.
8. Validation: The algorithm should have been validated and tested on relevant datasets to demonstrate its performance and reliability. Consider whether the algorithm has been benchmarked against other methods and evaluated in real-world applications.

By carefully considering these factors, researchers can select an algorithm that is well-suited to their specific application in bioinformatics or computational biology, ensuring that they obtain accurate, reliable, and meaningful results.

Widely Used Tools for Multiple Sequence Alignment

Clustal Omega

Clustal Omega is a multiple sequence alignment program that offers several features and capabilities for aligning biological sequences. Here are some key features of Clustal Omega:

1. Speed: Clustal Omega is designed for fast and efficient alignment of large datasets. It uses heuristics and algorithms optimized for speed without compromising alignment quality.
2. Scalability: Clustal Omega can handle large datasets with thousands of sequences and is scalable to accommodate increasing dataset sizes.
3. Accuracy: While optimized for speed, Clustal Omega also provides accurate alignments by incorporating advanced algorithms and heuristics.
4. Multiple Alignment Modes: Clustal Omega offers different alignment modes, including progressive alignment, which aligns sequences based on their pairwise similarities, and iterative refinement, which iteratively improves the alignment by realigning sequences.
5. Consistency-based Approach: Clustal Omega uses a consistency-based approach to improve alignment accuracy. It compares the alignment of each sequence pair to the alignment of other pairs to ensure consistency across the alignment.
6. Support for Various Input Formats: Clustal Omega supports a wide range of input formats for sequences, including FASTA, Clustal, and MSF formats.
7. Output Formats: Clustal Omega can generate output in various formats, including Clustal, MSF, and FASTA formats, making it compatible with other bioinformatics tools and software.
8. Web Interface and Command-line Version: Clustal Omega is available both as a web-based tool and a command-line version, providing flexibility for users to choose the interface that best suits their needs.
9. User-friendly Interface: The web-based version of Clustal Omega offers a user-friendly interface with options for adjusting alignment parameters and viewing the alignment results.
10. Open Source: Clustal Omega is open-source software, allowing users to access and modify the source code to suit their specific requirements.

URL: Clustal Omega

MAFFT

MAFFT (Multiple Alignment using Fast Fourier Transform) is a popular program for multiple sequence alignment (MSA) known for its speed, accuracy, and ability to handle large datasets. Here are some key features and capabilities of MAFFT:

1. Speed and Scalability: MAFFT is designed for speed and can efficiently align large datasets with thousands of sequences. It offers several strategies for accelerating alignment, including parallel processing and iterative refinement.
2. Accuracy: Despite its speed, MAFFT provides accurate alignments, especially for distantly related sequences. It uses progressive alignment followed by iterative refinement to improve alignment quality.
3. Multiple Alignment Methods: MAFFT offers several alignment methods, including FFT-NS-2 (Fast Fourier Transform-N-Strategy 2), which is the default method and is suitable for aligning large datasets, and L-INS-i (accurate method for large datasets and long sequences), which is more accurate but slower.
4. Support for Various Sequence Types: MAFFT supports various types of sequences, including nucleotide sequences (DNA and RNA) and protein sequences. It can also align sequences with large indels and size differences.
5. Multiple Output Formats: MAFFT can output alignments in various formats, including FASTA, Clustal, and PHYLIP formats, making it compatible with other bioinformatics tools and software.
6. Interactive and Batch Mode: MAFFT can be used interactively through a command-line interface or in batch mode for automated processing of multiple sequences.
7. Web Interface and Standalone Version: MAFFT is available as a standalone program for download and installation on local machines. It is also accessible through a web interface, making it convenient for users who prefer a graphical interface.
8. User-Friendly Options: MAFFT offers a range of options for customizing alignment parameters, allowing users to fine-tune the alignment process based on their specific requirements.
9. Continuous Development: MAFFT is actively developed and maintained, with updates and improvements regularly released to address user feedback and incorporate new features.

URL: MAFFT

T-Coffee

T-Coffee (Tree-based Consistency Objective Function for Alignment Evaluation) is a versatile multiple sequence alignment (MSA) program known for its ability to generate accurate alignments, especially for difficult cases such as sequences with low similarity or complex evolutionary histories. Here are some key features and capabilities of T-Coffee:

1. Multiple Alignment Methods: T-Coffee uses a combination of multiple alignment methods, including pairwise alignment, progressive alignment, and consistency-based alignment, to generate high-quality alignments.
2. Consistency-Based Approach: T-Coffee uses a consistency-based objective function to evaluate alignments, ensuring that the final alignment is consistent with the pairwise and multiple alignments of the input sequences.
3. Support for Various Sequence Types: T-Coffee supports the alignment of various types of sequences, including nucleotide sequences (DNA and RNA) and protein sequences. It can also handle sequences with large indels and size differences.
4. Integration of Structural Information: T-Coffee can incorporate structural information, such as protein secondary structure predictions or structural constraints, into the alignment process to improve accuracy.
5. Output Formats: T-Coffee can output alignments in various formats, including FASTA, Clustal, and PHYLIP formats, making it compatible with other bioinformatics tools and software.
6. Web Interface and Standalone Version: T-Coffee is available both as a standalone program for download and installation on local machines and as a web service through the T-Coffee server, providing flexibility for users to choose the interface that best suits their needs.
7. User-Friendly Interface: The web-based version of T-Coffee offers a user-friendly interface with options for adjusting alignment parameters and viewing the alignment results.
8. Advanced Options: T-Coffee offers advanced options for customizing the alignment process, allowing users to specify parameters such as alignment consistency scores and gap penalties.
9. Continuous Development: T-Coffee is actively developed and maintained, with updates and improvements regularly released to address user feedback and incorporate new features.

URL: T-Coffee

MUSCLE

MUSCLE (Multiple Sequence Comparison by Log-Expectation) is a widely used program for multiple sequence alignment (MSA) known for its speed, accuracy, and ability to handle large datasets. Here are some key features and capabilities of MUSCLE:

1. Speed and Scalability: MUSCLE is designed for speed and can efficiently align large datasets with thousands of sequences. It uses heuristics and algorithms optimized for speed without compromising alignment quality.
2. Accuracy: Despite its speed, MUSCLE provides accurate alignments, especially for distantly related sequences. It uses a progressive alignment approach followed by iterative refinement to improve alignment quality.
3. Multiple Alignment Methods: MUSCLE offers several alignment methods, including the default method, which is suitable for aligning large datasets, and the more accurate method, which is slower but provides better alignments.
4. Support for Various Sequence Types: MUSCLE supports various types of sequences, including nucleotide sequences (DNA and RNA) and protein sequences. It can also handle sequences with large indels and size differences.
5. Output Formats: MUSCLE can output alignments in various formats, including FASTA, Clustal, and PHYLIP formats, making it compatible with other bioinformatics tools and software.
6. Interactive and Batch Mode: MUSCLE can be used interactively through a command-line interface or in batch mode for automated processing of multiple sequences.
7. Web Interface and Standalone Version: MUSCLE is available as a standalone program for download and installation on local machines. It is also accessible through a web interface, making it convenient for users who prefer a graphical interface.
8. User-Friendly Options: MUSCLE offers a range of options for customizing alignment parameters, allowing users to fine-tune the alignment process based on their specific requirements.
9. Continuous Development: MUSCLE is actively developed and maintained, with updates and improvements regularly released to address user feedback and incorporate new features.

URL: MUSCLE

ProbCons

ProbCons is a probabilistic consistency-based algorithm for multiple sequence alignment (MSA) known for its accuracy, especially for aligning distantly related sequences. Here are some key features and capabilities of ProbCons:

1. Probabilistic Model: ProbCons uses a probabilistic model to guide the alignment process, incorporating information from both pairwise and multiple sequence alignments to improve alignment accuracy.
2. Consistency-Based Approach: ProbCons uses a consistency-based approach to evaluate alignments, ensuring that the final alignment is consistent with the pairwise and multiple alignments of the input sequences.
3. Accuracy: ProbCons is known for its accuracy, especially for aligning distantly related sequences. It uses a combination of probabilistic modeling and iterative refinement to improve alignment quality.
4. Support for Various Sequence Types: ProbCons supports various types of sequences, including nucleotide sequences (DNA and RNA) and protein sequences. It can also handle sequences with large indels and size differences.
5. Output Formats: ProbCons can output alignments in various formats, including FASTA and Clustal formats, making it compatible with other bioinformatics tools and software.
6. Web Interface and Standalone Version: ProbCons is available both as a standalone program for download and installation on local machines and as a web service through the ProbCons server, providing flexibility for users to choose the interface that best suits their needs.
7. User-Friendly Interface: The web-based version of ProbCons offers a user-friendly interface with options for adjusting alignment parameters and viewing the alignment results.
8. Advanced Options: ProbCons offers advanced options for customizing the alignment process, allowing users to specify parameters such as alignment consistency scores and gap penalties.
9. Continuous Development: ProbCons is actively developed and maintained, with updates and improvements regularly released to address user feedback and incorporate new features.

URL: ProbCons

Hands-On Exercise

Step 0 – installing Jalview

Go to https://www.jalview.org/, click the green downwards arrow and download the software for your platform. Install and test it. Note that the software starts showing a root window with several subwindows showing pre-loaded examples.

Part 1 – using MAFFT

One of the most widely used alignment programs is the “Clustal” package. It comes in two varieties: ClustalW (with a command line interface) and ClustalX (with a graphical interface). Typically, ClustalX is chosen for interactive use, and ClustalW is chosen when there is a need for automating the workflow. The underlying algorithm is precisely the same, and the results will be identical.

ClustalW/X is free to use and available for many computer systems — including Windows and Mac. If you want to install ClustalX on your own machine, it can be downloaded from here: http://www.clustal.org/clustal2/.

However, as mentioned during the Multiple Alignment lecture, ClustalW/X is neither the fastest nor the most precise program available. No single program performs best in all comparisons, but one of the best and fastest is called MAFFT. It can be downloaded from here: http://mafft.cbrc.jp/alignment/software/ (Note: command line interface only) or used online from here: http://mafft.cbrc.jp/alignment/server/. It is also available in an online version under EBI’s multiple alignment page: http://www.ebi.ac.uk/Tools/msa/.

Step 1

For the first part of the exercise we are going to consider a set of alpha-globin genes from a number of different animals. The first task is to construct a useful dataset. Below is a list of GenBank IDs for entries containing the sequences we need (some entries contain more than one gene).

GenBank: Nucleotide search

AB001981
X01831
AH002483 
J00043
J00044
X01086
X07053
AF098919

Open a text editor (e.g. jEdit) — as we find the genes we search for, we need to collect them in a FASTA file using descriptive short unique names.

NOTE:

1. By definition, FASTA names do not contain spaces, therefore use underscore or dash if you want to specify more than one word.
2. Names should be unique within the first 15 characters, since some programs (e.g. JalView) only consider the first 15 characters and fail in “interesting” ways if names are identical.
3. You should ignore the alpha-D-globin gene that has a CDS of only 89 nucleotides; it is not complete.
4. You should not include “embryonic alpha-type globin pi”.
5. There can be more than one alpha-globin gene in some of the GenBank entries. The resulting FASTA file should have 10 sequences.

For example, a sequence could be named like this:

>goat_alpha_globin_II
ATGGTGCTGTCTGCCGCCGACAAGTCCAATGTCAAGGCCGCCTGGGGCAAGGTTGGCAGCAACGCTGGAG
CTTATGGCGCAGAGGCTCTGGAGAGGATGTTCCTGAGCTTCCCCACCACCAAGACCTACTTCCCCCACTT
CGACCTGAGCCACGGCTCGGCCCAGGTCAAGGGCCACGGCGAGAAGGTGGCCGCCGCGCTGACCAAAGCG
GTGGGCCACCTGGACGACCTGCCCGGTACTCTGTCTGATCTGAGTGACCTGCACGCCCACAAGCTGCGTG
TGGACCCGGTCAACTTTAAGCTTCTGAGCCACTCCCTGCTGGTGACCCTGGCCTGCCACCACCCCAGTGA
TTTCACCCCCGCGGTCCACGCCTCCCTGGACAAGTTCTTGGCCAACGTGAGCACCGTGCTGACCTCCAAA
TACCGTTAA

For every GenBank entry find the genes (CDS features) coding for alpha-globin. We need the DNA sequence for the CDS features specifically — remember that you can click “CDS” to show that sequence only, and then change the display to FASTA format.

Copy each DNA sequence into your text editor as you find it. Give it a descriptive name that conveys which organism it comes from, and what type of alpha-globin it is. Remember to save your work often.

Important: Always use the gene/protein name from the CDS feature, not the GenBank entry name. (There is a trap buried somewhere here, where the entry name is directly misleading). See this PDF: How to locate CDS names in GenBank.

QUESTION 1:

Save your FASTA file and write the filename in your report. Hand in the file as an attachment to your report when you are finished.

Step 2

Go to EBI’s multiple alignment page: http://www.ebi.ac.uk/Tools/msa/ and choose the program MAFFT. Paste your sequences in the box (or upload the entire file). Choose “ClustalW” as Output format, and tell the program whether you are submitting Protein or Nucleic Acid sequences. Start the program.

Basically, MAFFT yields a purely text-based output, and all the graphics and tables you can see on the web page are added by EBI’s web server.

The tab “Alignment” shows the actual alignment.

QUESTION 2a:

1. What do you guess that the asterisks (“*“) under the alignment mean?
2. How many stretches of perfectly conserved sequence (of at least, say, 10 nucleotides) can you find? Write down the sequence(s) of the perfectly conserved stretch(es).

In the tab “Guide Tree” a graphical representation of the distances between the sequences is shown (note: this is not a “real” phylogenetic tree — it is an estimate based on the pairwise alignments. Actual phylogenetic trees need a multiple alignment as input — we’ll get there next week).

Look at the three main groups (clusters) that the sequences fall into.

QUESTION 2b:

1. Are the sequences “naturally” (biologically plausibly) placed? Or do the sequences seem to be randomly intermixed?
2. Do alpha-A and alpha-D seem closely or distantly related, sequence-wise?
3. What about alpha-1 and alpha-2?

It can be difficult to get a good overview over the sequences by looking at the raw text output. As an alternative, we will use the java-based “Jalview” program. Select the Results Viewers tab and click View result with Jalview. This is a link to a .jvl file, which should be opened with Jalview. The .jvl file contains a link to the alignment you just made. Note: if this does not work (on Linux, it doesn’t), open Jalview manually and then follow the instructions in the Results Viewers tab.

Jalview opens with a large root window, which contains a smaller alignment window. Note that both windows have their own menus. By default, the alignment is in black and white, but you can choose a colouring scheme in the Colour menu in the alignment window.

Explore the full length of the alignment in the JalView window by using the scrollbar in the bottom of the window — note the colouring of the nucleotides and the “consensus” line at the bottom. Note also that you can get information about the position of each nucleotide in each sequence by positioning the mouse over it.

QUESTION 2c:Include a screenshot of the alignment window in your report. It should show the 3′ part of the alignment (the rightmost part), and it should be coloured by nucleotide.

Step 3

Now translate the DNA sequences to protein sequences and construct a new alignment. Link: VirtualRibosome

QUESTION 3:

1. Insert the translated sequences in FASTA format in your report.
2. Examine the “Guide tree” tab again — do you get the same results as last time?
3. Inspect the alignment. How many perfectly conserved stretches can you find now (of at least, say, 5 amino acids)? Write down the sequence(s) of the perfectly conserved stretch(es).

Again, use “Jalview” to inspect the alignment (Note: if you want to launch Jalview via the .jvl file link, you have to close the program first. If you prefer to see both alignments at the same time, you should use the other option, i.e. File → Input Alignment → From URL).

• Experiment with different colouring schemes for the amino acids.
• Note also that a “conservation” and a “quality” score is calculated for each position.

Step 4

New data set: Insulin. This FASTA file contains the DNA sequence from the Insulin gene from a range of organisms

Notice that this FASTA file has been auto-generated from a database, and it is currently not that informative with regards to entry names. Before we carry on with the analysis, you need to figure out which organisms the sequences belong to by looking up the entries in GenBank. Based on this information construct a new FASTA file with names that 1) describe what organism each sequence came from and 2) keep in the GenBank ID for later reference.

Naming guideline: For example, the first entry (U00659 (now AH005355)) can be updated to:

>Sheep_AH005355
ATGGCCCTGTGGACACGCCTGGTGCCCCTGCTGGCCCTGCTGGCACTCTGGGCCCCCGCC
CCGGCCCACGCCTTCGTCAACCAGCACCTGTGCGGCTCCCACCTGGTGGAGGCGCTGTAC
CTGGTGTGCGGAGAGCGCGGCTTCTTCTACACGCCCAAGGCCCGCCGGGAGGTGGAGGGC
CCCCAGGTGGGGGCGCTGGAGCTGGCCGGAGGCCCCGGCGCGGGTGGCCTGGAGGGGCCC
CCGCAGAAGCGTGGCATCGTGGAGCAGTGCTGCGCCGGCGTCTGCTCTCTCTACCAGCTG
GAGAACTACTGTAACTAG

QUESTION 4:

• Use the naming guideline above to identify all species names, save your FASTA file and write the filename in your report. Hand in the file as an attachment to your report when you are finished.

Notice: As you might notice while you investigate the individual sequences in the FASTA file, it contains a certain level of redundancy (identical sequences). For now we keep in all the entries — we will learn from the multiple alignment we construct in the next step, which of the sequences are identical and should therefore be combined into a single entry.

Step 5

Construct a multiple alignment at the DNA level.

QUESTION 5:

1. Inspect the alignment (textual + JalView) – can you find any gap which is not a multiple of three (and which therefore cannot correspond to a number of whole codons)? Does it otherwise look like all gaps follow the codon boundaries?
2. By just eye-balling the differences between the sequences (easiest observed in JalView), can you immediately point to one sequence being the most “remote” (with the most differences compared to the rest)? Does this make taxonomical sense? (Hint: are all the sequences vertebrate?)
3. As mentioned above, some sequences are identical. Based on the guide tree – which of the sequences can we remove as being redundant (branch length = 0)?

Keep the browser window/tab with the DNA alignment open – we’ll need it for comparison in a moment.

Step 6

Now construct a multiple alignment at the peptide level.

• Once again look over the alignment — and pay special attention to the gaps, which will now truly represent the underlying codons. Try to see if you can find some of the locations that correspond to regions where small gaps had been inserted in the DNA alignment.

• Why do you think there may be a disagreement between the DNA and peptide alignment?

QUESTION 6:

1. Inspect the peptide alignment guide tree: Which of the sequences can we safely eliminate now? More than before?

Intermezzo – alternative splicing & protein isoforms

In the alignments we have been working with so far, the proteins have been related by evolution; either orthologous proteins from different organisms or paralogous proteins from the same organism (e.g. alpha-A and alpha-D globin). Now, we will work with a dataset of proteins related by alternative splicing: In some genes, introns can be spliced out of the transcript in more than one way, with the consequence that the same gene can produce a number of different proteins (isoforms). There are several kinds of alternative splicing, summarized in the figure to the right.

When aligning isoforms (proteins related by alternative splicing), it is important to realize that stretches of amino acid sequence are either completely identical (if they originate from the same stretch of nucleotide sequence) or completely unrelated (if they originate from different stretches of nucleotide sequence). Therefore, a correct alignment of isoforms will contain only matches and gaps, no mismatches.

Overview of different types of alternative splicing

Step 7

We are now going to investigate how three different alignment programs perform on a dataset of isoforms from one particular gene.

Here is a dataset consisting of 11 alternatively-spliced gene products from the human erythrocyte membrane protein band 4.1 (EPB). The goal of this exercise is to compare how well three different popular multiple alignment programs perform when attempting to align a set of proteins that are identical except for having different deletions.

Align the EPB sequences using the MAFFT, MUSCLE, and Kalign servers at the EBI Multiple Sequence Alignment page with Clustalw as output format. For each alignment, keep the window (or tab) open after use.

Now compare the the three alignments you just constructed using the EBI servers. For this purpose you should use the JalView alignment viewer linked on the EBI result pages. Remember that an alignment of isoforms ideally should have only matches and gaps, no mismatches.

QUESTION 7:

Are the three alignments different? Which, if any, of the three alignment methods got the alignment entirely correct?

You should note that this was just one particular form of test. On a different problem the relative performance of the alignment methods could well be different. However, you should also note that this was a fairly simple problem, and one where you could easily see artefacts. That will not be the case for most real biological data sets.

Part 2 – RevTrans

Step 8

As the final step in this exercise, we will have a look at how to get the best of both worlds: how to combine knowledge of both DNA and protein biology in a single multiple alignment (for the theory behind this, please refer to the RevTrans paper, linked on the main course page). RevTrans v.2 uses MAFFT as the default algorithm for constructing the peptide alignment — other options are ClustalW, T-Coffee and Dialign (a locally optimizing program, not available at the EBI servers).

If you haven’t read the RevTrans paper yet — please quickly skim through it now (it’s an easy read). The paper explains the concept behind the RevTrans method in details (DNA → Protein; Multiple alignment of the proteins; Construct DNA alignment from the DNA sequences using the peptide alignment as a scaffold).

The data set for this part of the exercise will be a cleaned-up version of the Insulin FASTA file from Step 4 above (redundancy reduced and with short informative names), available via this link.

Link to the RevTrans 2.0 server: https://services.healthtech.dtu.dk/service.php?RevTrans

Notice that it’s possible to specify which translation table to use. You may recognize the options from VirtualRibosome. This is no coincidence: both servers are using the same underlying algorithm for translating DNA to protein.

Paste in the sequences and start the RevTrans analysis with default settings.

QUESTION 8:

• Inspect the alignment:
  1. Are gap lengths always a multiple of 3?
  2. Are all codons aligned? (Codon 1st positions will be in the same columns as other 1st positions, 2nd positions only in columns with other 2nd positions etc.).

Closing remark: Currently the RevTRans server does not perform a lot of additional analysis on top of the actual alignment. The idea with the server is to provide input to “downstream” analysis in other tools. For example construction of phylogenetic trees and statistical analysis of silent vs. non-silent mutations (that is, mutations that do not change the amino acid sequence vs. mutations that do).