Abstract
Forensic DNA databases are a huge help in the investigation of crimes, making it possible to identify the person who left them using biological traces, provided that information about him in the form of an STR profile is already available. The same is true for unidentified corpses. When such information is not in the database, DNA phenotyping can come to the rescue, allowing you to restore the appearance of a person from his DNA, which is already used in forensic practice. The greatest progress has been made in establishing hair color, eye color, skin pigmentation and some other characteristics. But the main interest is the person’s face, and the situation with this is not yet the best, although there has been some progress. The main problem is that multiple genes are responsible for facial features, including a pleitropic effect. The advent of such a method as GWAS (Genome-Wide Association Study) made it possible to analyze many gene loci at once for the presence of single-nucleotide substitutions associated with certain genes involved in the formation of the human face. However, sequencing the two genomes (or exomes) of each person inherited from the father and mother with a phased haplotyped assembly of their sequences can become much more informative. And with this approach, it is necessary to correctly select objects in the form of a large number of doubles and their closest relatives, since, without being relatives, doubles can potentially carry the same nucleotide substitutions, which largely determine their external similarity. Another cohort should be families in which children are very similar to their parents, and in this case it is necessary to conduct triosequencing with phased assembly of their diploid genomes (exomes). The genetic information obtained in this way, processed using machine learning and artificial intelligence, can make it possible to “find” the desired genes, increasing the reliability of such DNA portraits.