Bioinformatics

Guide to AI and Data Courses · Part VI: Systems and Applications
Biological dataSequence alignmentGenome assemblyGene expressionProtein structureNeighbouring courses
Sercan KÜLCÜ · Giresun University · Computer Engineering

Introduction

What Question Does It Ask?

How are biological sequences and molecular data analyzed computationally?

Bioinformatics analyses biological data such as DNA, RNA and protein sequences with algorithms and statistics. The course covers problems such as sequence alignment, genome assembly, gene expression analysis and protein structure prediction, using methods that cope with the size and noise of this data.

At a glance

Approach

biology, algorithms and statistics

Input

DNA, protein sequences, expression data

Output

alignment, structure, function prediction

Success measure

biological validity

Math load

medium-high

Data needs

huge sequence datasets

Introduction

A Short History

1953

Watson and Crick, drawing also on Franklin's data, propose the double-helix structure of DNA.

1970

Needleman and Wunsch publish a global sequence alignment algorithm based on dynamic programming.

1981

Smith and Waterman publish the local sequence alignment algorithm.

1990

BLAST is published: fast, heuristic similarity search in large databases.

2003

The Human Genome Project is completed; sequencing costs begin to fall rapidly.

2020

AlphaFold 2 predicts protein structure at CASP14 with accuracy close to experimental methods.

The Course

Core Concepts

Sequence alignment

Global and local alignment; scoring matrices.

Sequence databases

GenBank, UniProt, PDB; similarity search.

Hidden Markov model

Gene finding and protein family profiles.

Genome assembly

Building a genome from short reads; de Bruijn graphs.

Gene expression

Measuring how active each gene is with RNA-seq.

Phylogenetics

Building the family tree of species from sequences.

Protein structure

Predicting 3D shape from the amino acid sequence.

Multiple testing

Controlling the false discovery rate over thousands of genes.

The Course

A Typical Workflow

SequencesequencingmachineClean andassembleQC, genomeassemblyAlignto reference,to databaseAnalysevariants,expressionInterpretbiologicalfunction
Example: A bacterium taken from a patient is sequenced, the short reads are assembled into a genome, and its genes are compared with known sequences in a database using BLAST. If antibiotic resistance genes are found, the physician is told which drugs will not work.

The Course

What Does the Course Cover?

  1. A molecular biology primer and biological databases
  2. Pairwise alignment: Needleman-Wunsch, Smith-Waterman
  3. Scoring matrices and similarity search with BLAST
  4. Multiple sequence alignment
  5. Hidden Markov models and gene finding
  6. Motif finding
  7. Genome assembly: de Bruijn graphs
  8. Read mapping and variant calling
  9. Phylogenetic trees
  10. Gene expression with RNA-seq; multiple testing
  11. Clustering and classifying gene expression
  12. Protein structure prediction and AlphaFold

Skills you will gain

handling sequences with Biopythonsimilarity search with BLASTread mapping with BWA and SAMtoolsexpression analysis with Bioconductorimplementing an alignment algorithm

A typical semester; the order and coverage vary between universities.

Try It

Align Two Gene Fragments

How similar are DNA fragments of the same gene in two species? The Needleman-Wunsch table finds the best alignment. Change the sequences and scores, then click a cell to see how it was computed.

Bioinformatics turns evolution into a scoring model: filling the table takes m×n steps, while trying every alignment is exponential. The same idea underlies tools such as BLAST.

Context

Prerequisites

Courses worth knowing before this one, and why they are needed.

CourseWhy it is needed
Machine Learninggene expression classification, protein structure
Data Miningsequence motifs, clustering
Algorithmsdynamic programming: the method behind sequence alignment
Data Structureshash tables, trees and graphs: k-mer counting and genome assembly
Biologythe basics of DNA, RNA, proteins and gene expression

Context

Are You Ready for This Course?

Ask yourself; for every "no", the right-hand column shows where to start.

Ask yourselfIf not, start with
Can you find the longest common subsequence of two strings by dynamic programming?Dynamic programming from an algorithms course
Can you explain how information flows from DNA to RNA to protein?High school biology or a short molecular biology primer
Can you count all substrings of length k in a text with a hash table?Hash tables from a data structures course
Can you train a classifier and interpret the result of a clustering?Machine Learning
If you can say "yes" to most of these, you are ready; one or two gaps can be closed during the semester.

Context

Neighbouring Courses

subfieldbuilds onshares methodsapplication

Overlap

Overlapping Courses

RelationCourseShared ground
applicationThis course applies the methods of Machine Learninggene expression classification, protein structure
applicationThis course applies the methods of Data Miningsequence motifs, clustering
shares methodsShares methods with Probabilistic Graphical Modelshidden Markov models for sequence alignment

Differences

Where It Differs (1/2)

CriterionBioinformaticsMachine LearningData Mining
Core questionHow are biological sequences and molecular data analyzed computationally?How does a program generalize from examples?How are hidden, useful patterns found in large piles of data?
Approachbiology, algorithms and statisticsstatistical learningdiscovery-driven analysis
InputDNA, protein sequences, expression datalabeled or unlabeled exampleslarge tables or transaction data
Outputalignment, structure, function predictiona model that makes predictionsrules, clusters, outliers
Success measurebiological validityerror on unseen datainterestingness, support, confidence
Math loadmedium-highhigh: linear algebra, probability, optimizationmedium
Data needshuge sequence datasetsmoderate to largelarge
  • Machine Learning: ML develops general-purpose learning methods; this course applies them to sequence and expression data together with biological knowledge.
  • Data Mining: Data mining looks for patterns in any data; this course uses alignment and motif methods specific to DNA and protein sequences.

Differences

Where It Differs (2/2)

CriterionBioinformaticsGraphical ModelsMedical Imaging
Core questionHow are biological sequences and molecular data analyzed computationally?How do we model and query many interdependent, uncertain variables?How do X-ray, CT and MRI images help diagnosis and treatment?
Approachbiology, algorithms and statisticsprobability and graphs combinedimage processing plus clinical knowledge
InputDNA, protein sequences, expression datavariables, dependencies, dataCT, MRI, X-ray, ultrasound
Outputalignment, structure, function predictionconditional probability, best explanationsegmentation, measurement, diagnostic suggestion
Success measurebiological validitylikelihood, inference accuracysensitivity, specificity, Dice
Math loadmedium-highhigh: probabilitymedium-high
Data needshuge sequence datasetsmoderatelittle; labels are expensive
  • Graphical Models: Probabilistic graphical models are a general framework; this course uses hidden Markov models for gene finding and protein families.
  • Medical Imaging: Medical image processing analyses images of organs and tissue; this course analyses sequence and expression data at the molecular level.

Practice

Applications

Personalized medicine

Choosing treatment by genetic variants.

Outbreak tracking

Tracing spread through virus genomes.

Drug discovery

Predicting target protein and molecule interactions.

Cancer genomics

Finding tumour-specific mutations.

Agriculture

Breeding productive and resilient crop varieties.

Evolution studies

Building species relationships from sequences.

Practice

Risks and Ethics

Genetic privacy

A genome identifies a person and their relatives; identities have been recovered from genetic data thought anonymous.

Ancestry-biased data

Genomic databases come mostly from people of European ancestry; risk scores are less accurate for other ancestries.

Batch effects

When samples are processed on different days or machines, a technical difference can look like a biological finding.

Clinical overinterpretation

Treating a variant of uncertain significance as a cause of disease can lead a patient to needless treatment or worry.

In depth: Explainable AI and AI Ethics.

Practice

Tools and Datasets

Tools and libraries

BLASTBiopythonBioconductor (R)BWASAMtoolsHMMERPyMOL

Well-known datasets

GenBankUniProtProtein Data Bank (PDB)Gene Expression Omnibus (GEO)1000 Genomes ProjectThe Cancer Genome Atlas (TCGA)

Practice

Common Misconceptions

Common beliefIn fact
One gene means one trait.Most traits come from many genes and the environment together; few are set by a single gene.
Once a genome is sequenced, its function is known.A sequence is only a string of letters; finding what genes do is a separate, hard analysis.
Bioinformatics is only for biologists.The core of the field is algorithms, data structures and statistics; computer engineers are central.
When thousands of genes are tested, the significant ones are real.Multiple testing inflates false positives; corrections such as the false discovery rate are needed.

Next Steps

After This Course

Courses that build on or extend this one

This course is mostly an end point: it builds on the others.

Related careers

BioinformaticianComputational biologistGenomics data scientistBiostatisticianAI researcher in drug discovery
Bioinformatics brings machine learning, data mining and probabilistic models together in one application area. After this course, computational genomics, structural biology and protein language models, or neighbouring health courses such as medical image processing, are natural next steps.

Next Steps

Further Reading

  • Durbin, Eddy, Krogh and Mitchison, Biological Sequence Analysis, Cambridge University Press, 1998.
  • Jones and Pevzner, An Introduction to Bioinformatics Algorithms, MIT Press, 2004.
  • Compeau and Pevzner, Bioinformatics Algorithms: An Active Learning Approach, 3rd ed., Active Learning Publishers, 2018.
  • Altschul et al., "Basic Local Alignment Search Tool", Journal of Molecular Biology 215(3), 1990.

The End

Chapter 39: Bioinformatics
← Previous chapter · Map of the field · Next chapter →
1 / 1 Sercan KÜLCÜ. All rights reserved.