This AI generates an "artificial genome": human DNA that does not belong to anyone

0 16

It is the first time that a human genome, albeit partial, has been created artificially using machine learning.

In 2016, the Human Genome Project-Write was controversial , because this initiative aims to create a complete human genome, but totally artificial; a stage of genetic engineering which raises ethical questions. In a paper published on February 4, 2021 in PLOS Genetics , another research team has just announced that it has succeeded in creating artificial human genomes, at least in part. It is a DNA that does not exist in itself, or in any case belonging to "imaginary humans", since it is entirely created by artificial intelligence.

The authors have indeed mobilized artificial neural networks in order to apply machine learning to genetic information. Artificial neural networks have been "  trained to learn the complex distributions of real genomic data sets  ", from human genomes accessible in data banks, so that they can then, for the first time in genetics, "  generate new high-quality artificial genomes  ”.

[bad iframe src]


The authors assure that the genomes produced are realistic: they are not simple reproductions of genomes from the training database, they are new. However, artificial genomes synthesize major aspects of the source database: for example, the allelic frequency (rate of presence of an allele in the genes of a population) is transcribed in the artificial genomes. It's as if the database is sort of the population that the genome came from - a mechanism similar to what happens in real life.

In their paper published in PLOS Genetics , geneticists demonstrate that artificial genomes produced with this technique retain many complex characteristics present in real genomes. "  For the major part of their properties, they are indistinguishable from other genomes in the biological databases that we used to train our algorithm, with the exception of one detail: they do not belong to any donor,  " commented authors.

This AI recreates most of the human genome from a machine learning training database. // Source: Burak Yelmen

Where the Human Genome Project-Write intends to synthesize a complete artificial genome in 10 to 20 years, the artificial genome presented by this other team in PLOS Genetics  could not reproduce a complete genome, but only a part of a genome. The limitations come mainly from the computing power required for such a computing process. But the authors explain that it would be possible to create a complete genome by producing each part separately and then combining them.

[bad iframe src]

This research remains at a preliminary stage, as "artificial genomes" may have many characteristics of real genomes, but our knowledge of genetic workings is far from complete, some differences may still escape - analysis must therefore be clarified. continue before concluding that this DNA is truly artificial human DNA, strictly speaking.


If the experience is impressive enough, the concrete utility can be questionable, just like the ethical issues that this poses.

The authors believe that artificial genomes may find utility in biomedical research, for which genetic data is a valuable resource, but rarely usable for privacy reasons. “  This creates a major scientific hurdle for researchers. Machine-generated genomes, or artificial genomes as we call them, can help us overcome the issue in a safe  , ethical framework , ”they write.

Genetic research could then be based on genetic data from imaginary humans, because an artificial genome only needs to exist for the machine learning algorithm to be able to extrapolate the biological characteristics resulting from this DNA (face, body, etc.) . According to these authors, this would therefore serve as a perfect replacement.

These geneticists, however, remain very evasive about the broader implications of such an achievement in terms of genetic engineering. Biological engineer Drew Endy said in the NY Times when announcing the Human Genome Project-Write in 2016: “  Before embarking on such a huge project, with such enormous ethical and theological implications, a question Basic ethics all the same need to be asked - starting with whether it is necessary, and under what conditions, to make these technologies a reality  ”.

$ 0.00