Unsupervised explainable AI for simultaneous molecular evolutionary study of forty thousand SARS-CoV-2 genomes

Abstract

Unsupervised AI (artificial intelligence) can obtain novel knowledge from big data without particular models or prior knowledge and is highly desirable for unveiling hidden features in big data. SARS-CoV-2 poses a serious threat to public health and one important issue in characterizing this fast-evolving virus is to elucidate various aspects of their genome sequence changes. We previously established unsupervised AI, a BLSOM (batch-learning SOM), which can analyze five million genomic sequences simultaneously. The present study applied the BLSOM to the oligonucleotide compositions of forty thousand SARS-CoV-2 genomes. While only the oligonucleotide composition was given, the obtained clusters of genomes corresponded primarily to known main clades and internal divisions in the main clades. Since the BLSOM is explainable AI, it reveals which features of the oligonucleotide composition are responsible for clade clustering. The BLSOM has powerful image display capabilities and enables efficient knowledge discovery about viral evolutionary processes.

Time-series trend of pandemic SARS-CoV-2 variants visualized using batch-learning self-organizing map for oligonucleotide compositions

“…Herein, we focused on pentanucleotide composition, but similar separations were obtained for other lengths of oligonucleotides (Ikemura et al 2020). BLSOM is an explanatory AI that can clarify combinatorial patterns of oligonucleotides that contribute to the separation according to clades and .…”

“…Next, it will be important to know the relationship between the strains isolated in clades and their subclusters and the causative mutations. When it comes to oligonucleotides as long as 15-mers, most are only present in one copy in the viral genome; therefore, changes in 15-mer sequences can be directly linked to mutations, and we have already started analysis from this perspective (Ikemura et al 2020).…”

“…We have previously revealed lineage-specific oligonucleotide compositions for a wide range of virus lineages and established a method to identify and classify viral-derived sequences in tick intestinal metagenomic sequences (Qiu et al 2019). In the case of SARS-CoV-2, we analyzed time-series changes in mono-and oligo-nucleotide compositions and found their time-dependent directional changes that are thought to be adaptive for growth in humans, which allowed us to predict candidates of advantageous mutations for growth in human cells (Ikemura et al 2020;Iwasaki Abe & Ikemura. 2021).…”

Unsupervised explainable AI for the collective analysis of a massive number of genome sequences: various examples from the small genome of pandemic SARS-CoV-2 to the human genome

“…Investigation of these black points revealed that they frequently contained the O (Other) clade sequences that GISAID did not classify as known specific clades. We believe that significant numbers of O-clade sequences can be classified into known clades (Ikemura et al, 2020). In Fig.…”

“…Due to the current explosive increase in available sequences, we must develop new technologies that can grasp the whole picture of big-sequence data and support efficient data mining. We have recently analyzed time-series changes in short and long oligonucleotide compositions in a large number of SARS-CoV-2 genomes and found many oligonucleotides that are expanding rapidly in the virus population, which allowed us to predict candidate advantageous mutations for growth in human cells (Wada et al, 2020b;Ikemura et al, 2020). Furthermore, the oligonucleotide BLSOM can classify the virus sequences into not only the known clades but also their subgroups (Abe et al, 2021).…”