Definition and a History of Bioinformatics as a Term (Video)
Written guide: Computational Biology in Tech
This video defines bioinformatics and traces the history of the term.
Chapters
- 0:00 What is bioinformatics
- 0:29 The NIH definition of bioinformatics
- 1:17 DNA and protein sequences as text
- 1:44 Sequencing reads and read alignment
- 2:34 Searching databases with BLAST
- 2:56 History of the word bioinformatics
- 3:53 Variant calling, RNA-seq, phylogenetics and AlphaFold
- 5:17 The SARS-CoV-2 genome in January 2020
- 7:30 How to start learning bioinformatics
- 8:06 Review
Video transcript
0:00 What is bioinformatics
Bioinformatics uses computers to store, search and interpret biological data, and modern biology depends on it because a single sequencing experiment produces far more data than anyone can read by hand. First, we will read a formal definition of the field, and then we will follow DNA from the sequencing instrument into a text file, go through the routine tasks, and trace one real example from the first weeks of the COVID-19 pandemic.
0:29 The NIH definition of bioinformatics
People use the word bioinformatics loosely, so a formal definition gives us a firm starting point. In July 2000, a committee at the US National Institutes of Health released a working definition of bioinformatics. In that document, bioinformatics covers the research, development or application of computational tools and approaches for expanding the use of biological, medical, behavioral or health data. The definition includes tools that acquire, store, organize, archive, analyze or visualize such data. Bioinformaticians develop these methods, sometimes as new software for searching and comparing biological data, and they apply them by running existing programs on new data. Now that we have the definition, we should ask what these data look like, and most of them are sequences.
1:17 DNA and protein sequences as text
DNA is a chain of nucleotides, and a protein is a chain of amino acids. Each nucleotide or amino acid has a one-letter symbol, so a computer can store a whole genome as a long string of text. The human genome is about three billion nucleotides long, and a computer holds all of those nucleotides as a string of letters.
1:44 Sequencing reads and read alignment
With sequences stored as text, the next question is how the letters get into the computer in the first place. Short-read sequencers determine the order of bases in DNA by reading it as millions of short fragments, which researchers call reads. A single read covers only a small stretch of the genome, so one experiment produces millions of reads to cover the full sequence. Once researchers have assembled a species' complete genome, that sequence becomes its reference genome. In read alignment, a program such as BWA finds the position in the reference from which each read came. Aligning reads resembles rebuilding a shredded page with an intact copy of the same book at hand, where the reader slides each strip along the printed page until the words line up.
2:34 Searching databases with BLAST
Alignment assumes we already know the genome that the reads came from, so we also need a way to identify a sequence that we do not recognize. Given a query sequence, which is the sequence we are asking about, BLAST finds similar sequences in public databases. The US National Center for Biotechnology Information maintains the program.
2:56 History of the word bioinformatics
Having met these two programs, we can place them in time. BLAST dates from 1990 and BWA from 2009, but the word bioinformatics is older. Paulien Hogeweg and Ben Hesper used it in a Dutch article in 1970 for the study of informatic processes in biotic systems, by which they meant the ways that living organisms handle information. Its meaning shifted in the late 1980s, when researchers began to use it mainly for computational methods that manage and analyze sequence data. The change followed the growth of the data. GenBank, the public sequence database of the United States, held 606 sequences in December 1982. By August 2026 it held approximately 267 million sequences, a count that leaves out the separate records from whole-genome sequencing projects.
3:53 Variant calling, RNA-seq, phylogenetics and AlphaFold
Alignment and searching are the two basic operations of the field, and most everyday analyses build on these operations. Variant calling compares aligned reads with the reference to find positions where a sample differs, such as a single-base change linked to a disease. RNA sequencing analysis counts how many reads come from each gene, which lets us compare gene activity between conditions such as tumor and normal tissue. Phylogenetics builds trees from aligned sequences, and in such a tree closely related organisms or virus strains sit on nearby branches. Sequence data also feed the prediction of protein shape. AlphaFold2, which Demis Hassabis and John Jumper at DeepMind presented in 2020, predicts the three-dimensional structure of a protein from its amino acid sequence. Its inputs are the sequence and an alignment of related sequences, and DeepMind trained the model on experimentally determined structures in the Protein Data Bank. More than 260 million predicted structures are now freely available in the AlphaFold database, which Google DeepMind developed with the European Bioinformatics Institute. Hassabis and Jumper shared half of the 2024 Nobel Prize in Chemistry for protein structure prediction.
5:17 The SARS-CoV-2 genome in January 2020
Now we can see how these methods work together on a real problem, in the first weeks of the COVID-19 pandemic. On 26 December 2019, the Central Hospital of Wuhan in China admitted a patient with a respiratory illness. Yong-Zhen Zhang's team at Fudan University in Shanghai sequenced the RNA in fluid washed from the patient's lungs and obtained more than 56 million reads. Consider how you would reassemble 56 million short fragments when the organism that produced them is unknown. With the cause unknown, the team had no reference genome for aligning the reads. Instead, they assembled the reads without a reference, using a program named Megahit that joins overlapping reads into longer stretches of sequence, which researchers call contigs. The longest contig was about 30,000 nucleotides long. When the team searched public databases with BLAST, the longest contig was closest to a coronavirus previously found in bats, at 89.1 percent nucleotide similarity. On phylogenetic trees, which the team built to show how the sequences were related, the new virus grouped with the SARS coronavirus and related viruses from bats. The team then shared the result openly. On 10 January 2020, they posted the genome sequence on virological.org, a public discussion forum for virologists. GenBank created a record for the genome two days later, and after two revisions, the second in March 2020, the record holds the complete genome of 29,903 nucleotides. The public sequence was put to use at once, because a team at the Charité hospital in Berlin had already aligned the genomes of related SARS-like viruses. As soon as the new sequence appeared, the team used their alignment to select PCR tests, which detect the virus by copying short stretches of viral genetic material. They designed the tests without access to the virus itself and published the tests on 23 January 2020.
7:30 How to start learning bioinformatics
Now that we have followed a complete example, we can consider what a beginner needs to learn. For a student who wants to analyze data from sequencing experiments, the foundations are the underlying biology and some statistics, along with a programming language, usually Python or R, and enough familiarity with the text-based command line to run programs. Computational biology is a neighboring field built more on mathematical models and simulation, and we compared the two fields in our video on bioinformatics and computational biology.
8:06 Review
Here is a short review of the main points. Computers store biological sequences as text, alignment places reads on a reference genome, and a database search finds the relatives of a query sequence. In January 2020, researchers used these same operations to identify a new virus from 56 million reads. Which kind of biological data would you most like to learn to analyze? Tell us in the comments.