Impact of using AI to generate primers for detection of SARS-CoV-2 Omicron variant

By Tarun Sai LomteReviewed by Aimee MolineuxFeb 1 2022

In a recent study posted to the bioRxiv* preprint server, researchers designed artificial intelligence (AI)-based primers for the severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) Omicron variant.

The Omicron (B.1.1.529) variant has more than 30 mutations in its genome that characterize it as the most divergent strain of SARS-CoV-2. While several mutations seen in Omicron have been detected in earlier variants, a combination of a high number of such mutations confer the variant with high transmissibility, increased binding affinity to the host cell receptor, immune evasion, decreased antibody neutralization among other properties.

*Important notice: bioRxiv publishes preliminary scientific reports that are not peer-reviewed and, therefore, should not be regarded as conclusive, guide clinical practice/health-related behavior, or treated as established information.

The identification of the Omicron variant was achieved using multiple targets or the search for gene target failures (S-gene target failure or SGTF), amplification, and sequencing. Although different methodologies exist to detect the Omicron variant in clinical samples, Omicron-specific primers are not available yet.

The study

In the present study, the authors developed Omicron-specific primers based on an AI technique. A method based on evolutionary algorithms (EAs) was employed to generate primer sets to identify Omicron sequences.

The sequence data of the Omicron and other variants were collected from the global initiative on sharing influenza data (GISAID) repository. About 123 sequences of the B.1.1.529 variant and 2100 sequences (100 each of different variants including the Alpha, Beta, Gamma, and Delta lineages) were stored in the FASTA format. Then authors included 100 sequences of the Omicron variant with those of other variants as the training set while the remaining Omicron sequences (23) formed the test set. 

Results

The authors obtained 10 different sequences after using the EA method 10 times. The team examined these sequences for the presence of more than one mutation and noted that one sequence (5’ GACCCACTTATGGTGTTGGTC 3’) contained three mutations (Q498R, N501Y, and Y505H) that are characteristic of the Omicron variant. This sequence was simulated in Primer3Plus as a forward primer (FP) and a “high end self complementarity” was observed that was resolved by the addition of a base pair (bp) at its end (5’ GACCCACTTATGGTGTTGGTCA 3’). An acceptable FP with a melting temperature (T_m) of 62 °C was obtained and subsequently, an internal probe (CACCAGCAACTGTTTGTGGA) and a reverse primer (3’ CTGCCAAATTGTTGGAAAGG 5’) were also generated with T_m of 60.8 °C and 60.5 °C, respectively.

Of the 123 Omicron sequences, only 112 sequences presented the FP sequence created by the researchers. Further, nine of the 11 Omicron sequences that did not present the FP sequence had errors in sequencing and two lacked the Y505H mutation. The sequence was tested with sub-lineages of the Omicron variant (BA.1, BA.2, and BA.3) to validate the findings. Further, the primers were tested in a laboratory to confirm that the primers identified the Omicron variant.

Conclusions

In the current study, the team successfully demonstrated the use of AI-based techniques to construct primers specific to the SARS-CoV-2 Omicron variant and its sub-lineages. Laboratory evaluation confirmed their AI-based results. The study findings, thus, could potentially make way for using AI-based solutions to create diagnostic tools to quickly respond to the ongoing coronavirus disease 2019 (COVID-19) pandemic and newly emerging SARS-CoV-2 variants in the future.

*Important notice: bioRxiv publishes preliminary scientific reports that are not peer-reviewed and, therefore, should not be regarded as conclusive, guide clinical practice/health-related behavior, or treated as established information.