French and Estonian researchers recently obtained artificial human genomes using artificial intelligence. These DNAs are unique and above all, they do not belong to anyone. What is the point of such research?
In 2017, we mentioned the Human Genome Project-Write, whose objective is to obtain a whole human genome, but completely artificial. At the time, ethical questions were already questioning this crucial step in genetic engineering . However, in a study published in the journal PLOS Genetics on February 4, 2021, geneticists from the University of Paris-Saclay (France) and the University of Tartu (Estonia) also claimed to have developed artificial human genomes. In reality, the researchers managed to obtain DNA only partially. However, the particularity of these works lies elsewhere. We are talking here about "non-existent" DNA, that is to say, not belonging to anyone, as if they were imaginary humans. These genomes were also designed by artificial intelligence.
Specifically, geneticists have used artificial neural networks and machine learning to process genetic information. This involved having the AI learn the complex distributions of real genomic datasets from human genome databases. Then the AI generated new artificial genomes , which is an astonishing first.
The scientists of the study are formal:the artificial genomes obtained are realistic and they are not reproductions of existing genomes. On the other hand, these retranscribe in part the source database , concerning for example the allelic frequency, relating to the rate of presence of an allele in the genes of a population. For the researchers, this technique makes it possible to preserve many complex characteristics found in authentic genomes. Finally, the only difference is that these genomes do not belong to any donor.
But what can these works be used for? According to geneticists, it is possible to find them useful in biomedical research . In this field, genetic data represent an important resource. On the other hand, they are rarely usable for confidentiality reasons. However, with artificial genomes, this problem is no longer a problem.
The research could therefore be based on genetic data from imaginary humans in order to allow an AI to imagine the biological characteristics related to these artificial genomes. However, the researchers did not provide an explanation about the possible broader applications of such a possibility in genetic engineering.
