Community structure and temporal dynamics of SARS-CoV-2 epistatic network allows for early detection of emerging variants with altered phenotypes (preprint)

The rise of viral variants with altered phenotypes presents a significant public health challenge. In particular, the successive waves of COVID-19 have been driven by emerging variants of interest (VOIs) and variants of concern (VOCs), which are linked to modifications in phenotypic traits such as transmissibility, antibody resistance, and immune escape. Consequently, devising effective strategies to forecast emerging viral variants is critical for managing present and future epidemics. Although current evolutionary prediction tools mainly concentrate on single amino acid variants (SAVs) or isolated genomic changes, the observed history of VOCs and the extensive epistatic interactions within the SARS-CoV-2 genome suggest that predicting viral haplotypes, rather than individual mutations, is vital for efficient genomic surveillance. However, haplotype prediction is significantly more challenging problem, which precludes the use of traditional AI and Machine Learning approaches utilized in most mutation-based studies. This study demonstrates that by examining the community structure of SARS-CoV-2 spike protein epistatic networks, it is feasible to efficiently detect or predict emerging haplotypes with altered transmissibility. These haplotypes can be linked to dense network communities, which become discernible significantly earlier than their associated viral variants reach noticeable prevalence levels. From these insights, we developed HELEN (Heralding Emerging Lineages in Epistatic Networks), a computational framework that identifies densely epistatically connected communities of SAV alleles and merges them into haplotypes using a combination of statistical inference, population genetics, and discrete optimization techniques. HELEN was validated by accurately identifying known SARS-CoV-2 VOCs and VOIs up to 10-12 months before they reached perceptible prevalence and were designated by the WHO. For example, our approach suggests that the spread of the Omicron haplotype or a closely related genomic variant could have been foreseen as early as the start of 2021, almost a year before its WHO designation. Moreover, HELEN offers greater scalability than phylogenetic lineage tracing methods, allowing for the analysis of millions of available SARS-CoV-2 genomes. Besides SARS-CoV-2, our methodology can be employed to detect emerging and circulating strains of any highly mutable pathogen with adequate genomic surveillance data.

From Alpha to Zeta: Identifying variants and subtypes of SARS-CoV-2 via clustering (preprint)

The availability of millions of SARS-CoV-2 sequences in public databases such as GISAID and EMBL-EBI (UK) allows a detailed study of the evolution, genomic diversity and dynamics of a virus like never before. Here we identify novel variants and sub-types of SARS-CoV-2 by clustering sequences in adapting methods originally designed for haplotyping intra-host viral populations. We asses our results using clustering entropy -- the first time it has been used in this context. Our clustering approach reaches lower entropies compared to other methods, and we are able to boost this even further through gap filling and Monte Carlo based entropy minimization. Moreover, our method clearly identifies the well-known Alpha variant in the UK and GISAID datasets, but is also able to detect the much less represented (< 1% of the sequences) Beta (South Africa), Epsilon (California), Gamma and Zeta (Brazil) variants in the GISAID dataset. Finally, we show that each variant identified has high selective fitness, based on the growth rate of its cluster over time. This demonstrates that our clustering approach is a viable alternative for detecting even rare subtypes in very large datasets.