![\"Do](\"https:\/\/www.newsbeep.com\/ie\/wp-content\/uploads\/2026\/03\/Do-Extreme-Environments-Shape.png\")
<\/p>\n <\/p>\n
\n Graphical abstract \u2014 University of Waterloo <\/p>\n
Evolutionary biologists have long known that DNA, the molecule that carries the genetic instructions for the development, functioning, growth and reproduction of all organisms, contains a record of ancestry. Indeed, the theory of common descent, a central pillar of modern evolutionary biology, holds that all life on Earth is related through an unbroken chain of genetic inheritance stretching back to a single ancestral organism in the distant past.<\/p>\n
Making sense of that web of evolutionary relationships relies increasingly on computation. Lila Kari, a Professor at the Cheriton School of Computer Science, uses mathematical and computational techniques to analyze DNA sequences, classify organisms and understand how they are related.<\/p>\n
But is ancestry the only story DNA tells? Could genomes also bear an imprint of the environments in which organisms evolved? According to a recent study, the answer, at least for some life forms and for some extreme environments, is an unexpected yes.<\/p>\n
The focus of the study is an unusual group of organisms known as microbial extremophiles, microscopic species that thrive at the edges of biological tolerance, in environments that by human standards are inhospitable if not outright lethal.<\/p>\n
\u201cExtremophiles are found in the harshest environments on Earth,\u201d Professor Kari says. \u201cThey\u2019ve been isolated from heated sediments near volcanoes, in deep-sea hydrothermal vents, in polar sea ice, almost every extreme environment biologists have examined for life.\u201d<\/p>\n
Some extremophiles, called hyperthermophiles, grow optimally at temperatures approaching 100\u00b0C. Others, known as psychrophiles, thrive at temperatures as low as -12\u00b0C. Still others inhabit highly acidic lakes with pH values near 0.5, highly alkaline lakes with pH around 11, or even solid rock several kilometres beneath the Earth\u2019s surface.<\/p>\n
Most extremophiles are bacteria and archaea, single-celled microbes so evolutionarily distant that they belong to different domains of life, the deepest division in the tree of life. Despite having diverged from one another about four billion years ago, researchers at the Cheriton School of Computer Science, working with colleagues at Western University and the University of Guelph, discovered that bacterial and archaeal extremophiles adapted to the same extreme environments can nonetheless exhibit strikingly similar genomic signatures.<\/p>\n
To understand how the research team detected this unexpected environmental signal in microbial genomic signatures requires a short digression into the structure of DNA, mathematical ways to represent DNA sequences, and different kinds of machine learning techniques used for identification and classification.<\/p>\n
Just as letters of the alphabet combine to form words, DNA molecules are composed of four \u201cgenetic\u201d letters \u2014 the nucleotides adenine (A), cytosine (C), guanine (G), and thymine (T) \u2014 linked together along the DNA phosphate backbone to comprise an organism\u2019s genome.<\/p>\n
One way to visualize patterns in the frequency and arrangement of these nucleotide letters is using a mathematical technique called Chaos Game Representation. In CGR, a DNA sequence is plotted as a two-dimensional image that captures how often specific nucleotide patterns occur.<\/p>\n
The illustration below shows how the short nucleotide sequence ACTCG is plotted (left), producing a simple image (right) that can be processed computationally.<\/p>\n
\u201cIf you plot longer DNA sequences, you find interesting, distinctive patterns,\u201d Professor Kari said. \u201cWhat\u2019s remarkable is that these CGR patterns are species-specific, which makes them useful as genomic signatures.\u201d<\/p>\n
Crucially, the patterns are preserved regardless of where the sequence is sampled within the genome, provided the fragment is long enough.<\/p>\n
\u201cThe first step was to choose a genome proxy for each extremophile, a DNA fragment long enough to represent the whole genome,\u201d Professor Kari said. \u201cIn our study, the optimal length was 100,000 base pairs, which is about three per cent of the average extremophile genome. We didn\u2019t take just one random segment, because we didn\u2019t want to introduce bias.\u201d<\/p>\n
To compare genomes visually and quantitatively, the team used a refined version of CGR called frequency Chaos Game Representation. Like CGR, FCGR represents a genome as a two-dimensional image, but it includes even more information: the frequencies of what are known as k-mers, sequences of DNA nucleotide letters of length k.<\/p>\n
For example, if the five-nucleotide sequence AGTCG appeared 20 times in a DNA fragment, that 5-mer would have a count of 20. From this k-mer data, a numerical vector of k-mer counts can be constructed for each DNA fragment, and it becomes the genomic signature unique to that organism. In an FCGR image, each pixel corresponds to a specific nucleotide sequence, and its intensity, how light or dark, reflects how often the k-mer count of that sequence appears in the genome.<\/p>\n
Using an empirically determined optimal proxy of 100,000 base pairs and k=6, the researchers generated FCGRs for all 693 extremophile microbes in the study.<\/p>\n
\u201cThe traditional view is that DNA contains only information about ancestry \u2014 taxonomy, who you are, and phylogeny, what you\u2019re related to,\u201d Professor Kari said. \u201cWe wanted to confirm that this signal was in our dataset.\u201d<\/p>\n
Using supervised machine learning, the team trained algorithms to classify DNA sequences by domain of life. The model was shown sequences labelled as bacterial or archaeal, learned the distinguishing patterns, and was then tested on new, unknown sequences.<\/p>\n
\u201cAs expected, the classification accuracy was extremely high, 99 per cent or better,\u201d Professor Kari says. \u201cThat confirmed that the strongest signal in a genome\u2019s k-mer profile is taxonomic.\u201d<\/p>\n
The researchers then asked a different question. Using the exact same genomic data, they retrained the algorithm, this time with environmental labels instead of taxonomic ones, and tested whether it could predict, say, whether an organism came from a hot or cold environment.<\/p>\n
\u201cThe accuracy wasn\u2019t as high as for taxonomy, but it was in the mid-70 per cent range,\u201d Professor Kari says. \u201cThat\u2019s far higher than random. It tells us some environmental information is embedded in these genomic signatures.\u201d<\/p>\n
To rule out the possibility that this result was an artifact of the model, the team then used unsupervised machine learning, which uses no prior information about taxonomy or environment.<\/p>\n
\u201cWith unsupervised learning, you remove all labels,\u201d Professor Kari explains. \u201cYou just give the algorithm the sequences and ask it to find what\u2019s similar to what for whatever reason.\u201d<\/p>\n
As expected, most clusters reflected evolutionary relationships, but a small number did not.<\/p>\n
\u201cSome clusters contained both bacteria and archaea,\u201d Professor Kari says. \u201cThat\u2019s remarkable. The difference between bacteria and archaea is fundamental. They shouldn\u2019t have similar genomic signatures at all.\u201d<\/p>\n
In total, the researchers identified 15 bacterial\u2013archaeal pairs that clustered together despite their immense evolutionary distance. When they examined where those organisms had been isolated, a startling pattern emerged.<\/p>\n
\u201cThey were found in similar extreme environments,\u201d Professor Kari says. \u201cThe clustering wasn\u2019t because of shared ancestry; it was because of shared environmental pressures. But the smoking gun is that for at least two pairs, not only were the bacterium and archaeon found in similar extreme environments but also in the same region geographically.\u201d<\/p>\n
One thermophilic, acidophilic bacterium\u2013archaeon pair was found co-existing in Washburn Hot Springs in Yellowstone National Park. Another hyperthermophilic pair co-occurred in not only one, but two different locations: Brothers Volcano, a submarine volcano near New Zealand, and the Juan de Fuca Ridge, an underwater mountain range off the coast of Vancouver Island.<\/p>\n
\u201cWe found that extreme environments can stamp a powerful, genome-wide watermark on DNA,\u201d Professor Kari said. \u201cEven microbes from the two most distant branches of life \u2014 bacteria and archaea \u2014 can end up with strikingly similar genomes, and in some cases we found these unlikely partners living side-by-side in the very same place.\u201d<\/p>\n