Introduction
Since the onset of the COVID-19 outbreak in Wuhan, China, in November 2019, the global toll has reached approximately 7 million deaths and 700 million confirmed cases [Reference Wu1, 2]. While mortality rates have declined, the virus continues to pose significant risks, particularly for individuals with underlying health conditions and those who remain unvaccinated [Reference Ghafari3, Reference Spinner4]. In Hong Kong, there have been 2.9 million reported cases and over 14 000 fatalities due to COVID-19 [5]. The case fatality rate (CFR) for Hong Kong stands at 0.508%, which is lower than those in China (1.047%), the United Kingdom (0.932%), and the United States (1.09%). However, this rate is higher than those reported by neighbouring countries, including South Korea (0.104%), Japan (0.221%), and Taiwan (0.186%) [Reference Yang6].
The public health response in Hong Kong has been dynamic, with policies evolving throughout the pandemic. The efficacy of these measures is the subject of a recent work by Yang et al., in which 4 of 6 epidemic waves from 2020 to 2022 were contained, and case-fatality risk was reduced through pharmaceutical and non-pharmaceutical interventions [Reference Yang6]. On 5 May 2023, the World Health Organization (WHO) officially declared an end to the global public health emergency response to the COVID-19 pandemic, citing widespread vaccination efforts that have led to reduced mortality, hospitalizations, and ICU admissions [7]. Hence, surveillance of SARS-CoV-2 has decreased significantly over the past year due to declining case numbers, perceived herd immunity, and pandemic fatigue.
Despite these developments, the ongoing circulation of SARS-CoV-2 poses a continued risk of viral evolution. Mutations may emerge that confer advantages such as increased transmissibility, immune evasion, enhanced fitness, replication capacity, virulence, or resistance to existing interventions [Reference Pilapil, Notarte and Yeung8, Reference Fokam9]. Vigilance and ongoing research are essential for monitoring and responding to potential changes in the virus.
The COVID-19 pandemic has highlighted the critical importance of understanding the evolutionary dynamics of SARS-CoV-2 to guide effective public health responses. Genomic surveillance has been vital in detecting and monitoring the emergence of new variants, providing crucial insights into their dissemination and impact. This information is essential for informing vaccination strategies, therapeutic advancements, and public health directives. The identification of key mutations has demonstrated the virus’s capacity for rapid genetic diversification and adaptation, underscoring the need for continuous monitoring and research.
In this study, we conducted a comprehensive analysis of SARS-CoV-2 genome sequences collected in Hong Kong. Our objectives include delineating the virus’s genetic diversity, identifying significant mutations and lineages, and assessing their epidemiological and public health implications. By analysing the binding affinities of various spike protein mutations to the ACE2 receptor, we aim to elucidate the underlying changes in viral entry and transmission. Additionally, we explored the influence of demographic factors, such as age and sex, on the evolution of SARS-CoV-2 infections.
Our study aims to provide a long-term description of the evolutionary dynamics of SARS-CoV-2 in Hong Kong. The integration of genomic surveillance and detailed mutation analysis enhances our understanding of the impacts of viral evolution on public health. Such efforts facilitate effective monitoring and control of the virus and contribute to a better understanding of SARS-CoV-2’s underlying transmissibility.
Methods
Study design and subsampling
As of 12 January 2024, the NCBI and GISAID repositories contained approximately 8.5 million and 16.4 million SARS-CoV-2 sequences, respectively [Reference Brister10, Reference Khare11]. Among the genomes deposited between 21 January, 2020 and 9 December 2023, we identified 2 520 sequences from NCBI and 19 457 sequences from GISAID that originated from Hong Kong, after excluding duplicate entries (n = 2 524). We subsequently selected sequences of human origin (n = 19 429) with read lengths ≥29 000 base pairs (n = 18 558) and those accompanied by collection dates (n = 15 982) for lineage classification and mutation profiling.
Cross‑repository duplicates were defined by identical consensus sequence (length ≥ 29 000 bp; exact match by hash), matching collection date and location, and concordant metadata. For duplicate sets, we retained the record with the highest sequence quality (lowest ambiguity, greatest coverage) and most complete metadata. We then excluded sequences with coverage <90% or > 5% ambiguous bases and resolved metadata inconsistencies (e.g., invalid dates) before standardizing lineage assignment.
Additionally, we also obtained SARS-CoV-2 sequences from Guangdong for comparison with those from Hong Kong following the reopening of the border between the two regions. A total of 6 358 sequences from Guangdong deposited in NCBI and GISAID from 10 January 2020 to 30 June 2023, were included in this analysis.
Phylogenetic analysis of SARS-CoV-2 in Hong Kong
We aligned a subsample of 15 982 SARS-CoV-2 sequences from Hong Kong to the reference genome Wuhan-Hu-1 (GenBank accession no. MN908947.3). Only genomes with ≥90% coverage and < 5% ambiguous nucleotides were retained. Maximum likelihood phylogenetic inference was performed using IQ-TREE [Reference Nguyen12]. The resulting phylogenetic tree was refined with TreeTime [Reference Sagulenko, Puller and Neher13], which involved rerooting, resolving polytomies, inferring nodes, and pruning sequences. Mutations were reconstructed by inferring nucleotide changes and translating them into amino acids. Clade and lineage classifications followed the PANGO [Reference Rambaut14, Reference O’Toole15] and Nextstrain [Reference Hadfield16] nomenclatures.
Correlation analysis
We selected sequences that included sex and age information, resulting in 6 052 and 4 671 data points, respectively. To preserve statistical power and avoid further reducing the sample size, sex and age were analysed separately. While only a subset (n = 4 671) included both variables, this trade-off allowed for a broader assessment of variant–demographic associations, which aligns with the descriptive aims of this study. The contribution of each variant was summed, and we conducted Chi-squared tests for independence and Pearson’s correlation analysis (p ≤ 0.05) using R. Given that the samples were classified into categorical variables, only variants with normal Pearson residuals were plotted.
Calculation of binding affinity of SRBD to ACE2
The spike protein sequence of SARS-CoV-2 Wuhan-Hu-1 (UniProt ID: P0DTC2) and the spike receptor-binding domain-angiotensin-converting enzyme 2 (SRBD-ACE2) complex structure (PDB ID: 6M0J) were used as references. Mutations from SARS-CoV-2 variants of concern were incorporated into the spike protein sequence, and structure prediction was performed using AlphaFold2 [Reference Jumper17]. The models were visualized and inspected, then energy-minimized using PyMOL [Reference Schrödinger18], ChimeraX [Reference Meng19], and Yasara [Reference Krieger20]. Quality assessment was conducted using reference PDBs and Ramachandran plots [Reference Berman21].
The binding scores of the SRBD from multiple SARS-CoV-2 variants to the ACE2 receptor were calculated through molecular docking simulations based on experimentally identified active residues [Reference Lan22]. For each variant, ten replicates of binding scores were generated to represent different protein complexes, and the mean binding score and standard deviation were then calculated. Statistical analyses, including One-way Analysis of Variance, Tukey’s pairwise comparisons, and Hedges’ g effect sizes, were conducted to quantify the mean binding scores among the variants. These analyses were performed using the Pandas and NumPy libraries in Python [Reference Van Rossum and Drake23–Reference Harris25]. Visualization of the SRBD–ACE2 interaction was created using BioRender [Reference Ryan and Caplice26].
Results
SARS-CoV-2 variants and mutations throughout the pandemic
A total of 543 lineages and 5 982 unique mutations were identified from 15 982 quality-controlled SARS-CoV-2 genome sequences collected from Hong Kong between 21 January 2020 and 09 December 2023 (Figures 1 and 2; Supplementary Tables S2-S4). The results are organized by waves of infection, highlighting the emergence and prevalence of various lineages and mutations over time.
Temporal density plots of the ten most abundant SARS-CoV-2 lineages in Hong Kong (2020–2023). This figure presents the density plots of SARS-CoV-2 lineages from 2020 to 2023, illustrating their temporal circulation. The x-axis represents time, while the y-axis shows the density of each lineage’s presence. Each curve, generated using kernel density estimation (KDE), reflects the relative prevalence of a variant over time, rather than its overall average. The colour-coded curves represent each lineage’s trends, highlighting shifts in circulation patterns. The Wuhan-Hu-1 strain dominated the first wave (January–February 2020), the second wave (March–April 2020) by the B and B.1 variants, the third wave (July–September 2020) by the B.1.1.63 variant, the fourth wave (November 2020–April 2021) by the B.1.36 variant, the fifth wave (January–April 2022) by the BA.2 subvariants, and the sixth wave (May–October 2022) by BA.2.2 and BA.5 subvariants.
Figure 1 Long description
Temporal Density Plots of the Ten Most Abundant SARS-CoV-2 Lineages in Hong Kong (2020–2023). This figure presents density plots of SARS-CoV-2 lineages from 2020 to 2023, illustrating their temporal circulation. The x-axis represents time, while the y-axis shows the density of each lineage’s presence. Each curve, generated using kernel density estimation (KDE), reflects the relative prevalence of a variant over time, rather than its overall average. The color-coded curves represent each lineage’s trends, highlighting shifts in circulation patterns. The Wuhan-Hu-1 strain dominated the first wave (January–February 2020), the second wave (March–April 2020) by the B and B.1 variants, the third wave (July–September 2020) by the B.1.1.63 variant, the fourth wave (November 2020–April 2021) by the B.1.36 variant, the fifth wave (January–April 2022) by BA.2 subvariants, and the sixth wave (May–October 2022) by BA.2.2 and BA.5 subvariants.
Prevalent amino acid substitutions and deletions in the spike protein of SARS-CoV-2 variants. This figure summarizes the most prevalent amino acid substitutions and deletions in the spike protein of selected SARS-CoV-2 variants of concern: (a) N-terminal domain (residues 31–305), (b) Spike receptor-binding domain (residues 319–541), and (c) Remaining S1 and S2 subunits (residues 542–1 273). The heatmap visualizes the prevalence of these mutations as percentages (0%–100%), with intensifying shades of blue representing every 20% increment. This summary of predominant mutations per lineage served as the basis for the SRBD-ACE2 simulation.
Figure 2 Long description
Prevalent Amino Acid Substitutions and Deletions in the Spike Protein of SARS-CoV-2 Variants. This figure summarizes the most prevalent amino acid substitutions and deletions in the spike protein of selected SARS-CoV-2 variants of concern: A) N-terminal domain (residues 31–305), B) Spike receptor-binding domain (residues 319–541), and C) Remaining S1 and S2 subunits (residues 542–1273). The heatmap visualizes the prevalence of these mutations as percentages (0-100%), with intensifying shades of blue representing every 20% increment. This summary of predominant mutations per lineage served as the basis for the SRBD-ACE2 simulation.
First wave (January–February 2020)
The earliest complete SARS-CoV-2 genome sequences were documented from two male patients aged 39 and 56 on 21 January 2020, marking the onset of the first wave, which lasted until February 2020 (Figure 1a). During this period, daily cases remained below ten, with 135 sequenced cases recorded out of a total of 1 822 sequences deposited for the year (Supplementary Figures S1 and S2). The predominant lineages were A (0.44%, n = 8), B (6.81%, n = 124), and B.4 (0.17%, n = 3). Key mutations from the 135 sequenced cases included G251V (72.59%, n = 98) in ORF3a, H3233Y (40.74%, n = 55) in ORF1a, and L8V (42.22%, n = 57) in the spike (S) protein.
Second wave (March–April 2020)
The second wave saw an increase in diversity, with 18 lineages among the 1 822 sequences deposited that year (Figure 1a). Notable lineages included B.1 (2.63%, n = 48), B (2.20%, n = 40), B.1.1 (1.65%, n = 30), B.3 (1.21%, n = 22), and B.1.576 (0.82%, n = 15). The D614G mutation was present in 55.14% of cases, while P314L was found in 52.97% of sequences. In May and June 2020, lineage diversity decreased, with B.1.576 (0.60%, n = 11) and B.1 (0.38%, n = 7) remaining prevalent, both exhibiting the D614G and P314L mutations (75.76%).
Third wave (July–September 2020)
The third wave peaked in July 2020, with over 100 confirmed cases daily, and lasted until September 2020 (Figure 1a). A total of 28 lineages were identified, with B.1.1.63 (31.23%, n = 569), B.1.1 (1.76%, n = 32), and B.1 (1.54%, n = 28) being the most common. The D614G mutation was present in 98.87% (n = 696) of cases, along with P314L in 98.29% (n = 692). Seven additional mutations were detected, including R203K (87.78%, n = 618), G204R (87.78%, n = 618), and A12G (77.13%, n = 543) in the nucleoprotein (N), S12F (79.55%, n = 560) in the S protein, A903V (78.41%, n = 552) in ORF1a, A2282V (75.28%, n = 530) in ORF1b, and H9D (77.13%, n = 543) in ORF9b.
Fourth wave (November 2020–April 2021)
A resurgence of COVID-19 cases began in November 2020, peaked in December, and continued until April 2021 (Figure 1a,b). This wave was primarily driven by the B.1.36 variant (56.20%, n = 1 024). Other variants included Beta (B.1.351, 3.68%, n = 67) and Alpha (B.1.1.7, 3.35%, n = 61). Delta variants B.1.617.2 (1.32%, n = 24) and AY.75 (0.55%, n = 10) and Kappa (B.1.617.1, 1.37%, n = 25) were also present. Six new predominant mutations were found, including S194L (78.50%, n = 1 033) in the N protein, the V561A (75.30%, n = 991), V1056L (74.92%, n = 986), and A1997V (74.62%, n = 982) in ORF1a, and Q57H (85.48%, n = 1 125) and T223I (74.32%, n = 978) in ORF3a.
From May to December 2021, local transmission of COVID-19 in Hong Kong was nearly nonexistent, with daily case counts consistently remaining below 20. During this period, a total of 924 sequenced cases were recorded out of 1 545 sequences deposited for the year. The dominant variants during this time were Delta variants, including B.1.617.2 (7.25%, n = 112), AY.4 (3.75%, n = 58), AY.43 (3.69%, n = 57), AY.122 (3.56%, n = 55), and AY.127 (2.91%, n = 45). Additionally, Omicron lineages began to emerge, with significant representation from BA.1 (5.37%, n = 83), BA.1.1 (5.89%, n = 91), BA.1.17.2 (2.91%, n = 45), BA.2.3 (1.42%, n = 22), and BA.1.17 (1.10%, n = 17).
Fifth wave (January–April 2022)
The fifth wave began in January 2022, peaked between February and March, and concluded in April 2022 (Figure 1c). This wave was mainly attributed to Omicron variants, particularly BA.2.2 (24.55%, n = 2 804), BA.2 (1.52%, n = 174), and BA.1 (1.18%, n = 135). Other variants include BA.1.1 (3.53%, n = 131), AY.127 (2.91%, n = 108), BA.2.3 (2.77%, n = 103), and BA.2.10 (1.15%, n = 51). Omicron lineages exhibited a higher mutation rate compared to previous variants.
Sixth wave (May–October 2022)
From May to October 2022 (Figure 1c), the BA.2.2 variant dominated (26.85%, n = 3 067), alongside BA.5 descendants, including BA.5.2 (7.94%, n = 907), BA.5.2.1 (3.69%, n = 422), and BA.5.6 (3.17%, n = 362). This period marked a slight increase in mutation rates.
2023 Variants and trends
In the first half of 2023, lineages such as CM.12 (6.88%, n = 82), CH.1.1.11 (5.29%, n = 63), FL.4 (4.95%, n = 59), and XBB.1.16 (4.87%, n = 58) were predominant (Figure 1d). The reopening of borders with Mainland China on 8 January 2023 [27] saw local circulating variants remain distinct from those in Guangdong. The locally circulating variants in Hong Kong included CH.1.1 (8.64%, n = 103), CM.12 (6.80%, n = 81), BA.5.2 (4.61%, n = 55), and BQ.1.1 (2.18%, n = 26). These variants are different from the prevalent DY.3 lineage in the neighbouring Guangdong province, which accounted for 40.88% (n = 491) of cases, as well as from the co-circulating variants BF.7.14 (10.49%, n = 126), DY.4 (7.66%, n = 92), and DY.2 (6.99%, n = 84).
Following the WHO’s declaration on 5 May 2023, which marked the end of the COVID-19 global health emergency, the variant composition in both Hong Kong and Guangdong changed significantly. Notably, the emergence of lineages FL.2.3 (16.55%, n = 589) and XBB.1.16 (14.78%, n = 526) was observed in both regions. This shift indicates the inevitable transmission of SARS-CoV-2 between the two areas, facilitated by unrestricted cross-border travel.
Mutation frequency and transmissibility
The frequency of mutations varied among strains relative to the Wuhan-Hu-1 strain. From the beginning of the pandemic until the end of the fifth wave, the mean amino acid substitutions were 22.28 ± 2.63 for Alpha, 18.87 ± 1.72 for Beta, 33.92 ± 4.63 for Delta, and 52.00 ± 3.09 for Omicron. Omicron lineages continued to evolve, with mean substitutions of 54.51 ± 2.75 during the sixth wave, rising to 73.94 ± 3.46 in the second half of 2023.
In terms of transmissibility, 99.23% (n = 1 808) of sequences in 2020 were ancestral variants, with Alpha and Beta variants present at low frequencies (0.66% and 0.11%, respectively). By early 2021, the Alpha variant accounted for 4.27% (n = 66) of sequenced cases in Hong Kong. In contrast, its prevalence was significantly higher in neighbouring regions, with 10.84% in Guangdong, 21.66% in the Philippines, and 74.92% in Taiwan [Reference Khare11]. These figures suggest that the Alpha variant entered Hong Kong later than in these other areas. The Delta emerged as the dominant variant later that year, accounting for 40.71%. Omicron appeared towards the end of 2021, rapidly increasing to 98.63% (n = 11 266) of sequences in 2022 and maintaining a prevalence of 97.40% (n = 1 161) in 2023, with recombinant variants representing 2.60% (n = 31) of sequences. This surge in Omicron variants was consistent with global trends, driven by factors such as high transmissibility and the relaxation of public health measures.
SARS-CoV-2 prevalence by age and sex
The prevalence of COVID-19 cases showed notable differences across age groups and sex in Hong Kong (Figure 3). Among the sequenced cases with available patient age and sex data, the absolute proportion was highest in the 30–39 age group (17.8%), followed by 40–49 (14.7%), 20–29 (13.8%), and 80+ (12.6%). The lowest proportions are seen in the youngest age groups, 0–9 (4.4%) and 10–19 (5.1%). This distribution highlights that working-age adults and the elderly are more frequently represented among COVID-19 cases in this dataset, while children and teenagers make up a smaller share. Such patterns may reflect differences in exposure risk, social behaviour, occupational roles, and possibly the impact of public health interventions or vaccination campaigns.
Bar graphs of the COVID-19 cases distribution across age groups and sex. This figure presents bar graphs illustrating the distribution of sequenced COVID-19 cases: (a) Absolute proportion of COVID-19 cases across age groups, (b) Normalized distribution of COVID-19 cases across age groups (per 1 000 population), (c) Absolute proportion of COVID-19 cases across sex and age group combinations, and (d) Normalized distribution of COVID-19 cases across sex and age group combinations (per 1 000 population).
Figure 3 Long description
Bar Graphs of the COVID-19 Cases Distribution Across Age Groups and Sex. This figure presents bar graphs illustrating the distribution of sequenced COVID-19 cases: A) Absolute proportion of COVID-19 cases across age groups, B) Normalized distribution of COVID-19 cases across age groups (per 1 000 population), C) Absolute proportion of COVID-19 cases across sex and age group combinations, and D) Normalized distribution of COVID-19 cases across sex and age group combinations (per 1 000 population).
In terms of the combined age–sex proportions, a nearly balanced distribution between males and females was observed in 0–9, 10–19, 50–59, and 80+ age groups. Among working-age adults (20–49), females make up a higher proportion of cases, especially in the 30–39 and 40–49 age groups. In older age groups (60–79), the distribution shifted, with males becoming more prevalent. These patterns reflect a combination of biological, social, and behavioural factors. For instance, the higher proportion of female cases in certain adult age groups may be linked to occupational exposure, particularly in healthcare and caregiving roles. In contrast, the male predominance observed in older age groups might be influenced by factors such as differences in susceptibility, comorbidities, or health-seeking behaviour.
To elucidate temporal trends in SARS-CoV-2 epidemiology in Hong Kong, we examined the normalized distribution of all sequenced COVID-19 cases (cases per 100 000 population) across age-sex strata in six-month intervals from January 2020 through December 2023. This approach offers a granular perspective on shifting risk profiles amid evolving public health policies, viral dynamics, and population immunity.
January–June 2020
During the initial six months of the pandemic, COVID-19 incidence was low (3.42 per 100 000; 256 cases among approximately 7.48 million), with modest elevations among young working-age adults relative to children and many older-adult strata, and minimal sex differences (Figure 4). For instance, young-adult females had per-100 000 values around 3.4–5.0 in 20–39, and males in 20–39 around 4.7–6.9; children were notably lower (e.g., Female 0–9 roughly 0.75 per 100 000; Male 0–9 roughly 1.41 per 100 000). This profile is consistent with rapid border tightening, compulsory quarantine of inbound travellers, centralized isolation of confirmed cases, and high-performing contact tracing that constrained community spread, limited spillover into seniors, and left incidence concentrated among travelling-age adults and essential workers.
Bar graphs of the COVID-19 cases distribution across age groups and sex throughout the pandemic. This figure presents bar graphs illustrating the normalized distribution of sequenced COVID-19 cases (per 100 000 population) every six months throughout the pandemic (January 2020 to December 2023).
Figure 4 Long description
Bar Graphs of the COVID-19 Cases Distribution Across Age Groups and Sex Throughout the Pandemic. This figure presents bar graphs illustrating the normalized distribution of sequenced COVID-19 cases (per 100 000 population) every six months throughout the pandemic (January 2020 to December 2023).
July–December 2020
Semiannual incidence increased to 12.12 per 100 000 (900 cases among approximately 7.43 million), with males modestly higher than females (12.96 compared to 11.41 per 100 000). The age gradient shifted upward, with elevated incidence among older working-age and early-elderly groups compared with paediatric strata. Specifically, 70–79 years reached 18.44 per 100 000 and 60–69 years with 16.35 per 100 000, while 30–39 and 40–49 years were 12.32 and 11.92 per 100 000, respectively; paediatric rates remained lower (0–9 at 4.72; 10–19 at 7.03 per 100 000). Despite episodic community, workplace, and healthcare-associated clusters, magnitudes remained far below the long-term care facility (LTCF)-amplified senior burdens seen in North America and Europe in 2020, highlighting the effectiveness of targeted protections in care settings and rapid outbreak control. However, the elevation in the 60–79 age group emphasized the vulnerability of older cohorts even under stringent PHSMs.
January–June 2021
Aggregate semiannual incidence in 2021 was low to moderate compared with 2020 peaks in Hong Kong, with waves largely constrained by continued test–trace–isolate protocols, quarantine, and stringent travel controls. Initial vaccine rollout (Comirnaty and CoronaVac) began, but uptake was slower among the oldest groups [Reference Yuan28, Reference Xiao29]. Compared globally, many locations experienced recurrent waves driven by Alpha and Delta, but Hong Kong maintained suppression, resulting in fewer local chains and a lower overall incidence than in Southeast Asian neighbours, Western Europe, or the USA during the same period [Reference Cahyani30–Reference Walensky, Walke and Fauci32]. Age distribution remained working-age adult-skewed relative to children, consistent with intermittent school measures and limited community mixing, though not as senior-weighted as later Omicron.
July–December 2021
Semiannual incidence was 7.67 per 100 000 (568 cases among approximately 7.40 million). Males exceeded females (10.24 compared to 5.52 per 100 000), consistent with occupational exposures, mobility patterns, and testing requirements. Incidence remained highest in young adults (20–29 at 17.06; 30–39 at 11.33 per 100 000), while paediatric and senior rates were comparatively low.
With borders largely closed and case-based controls in place, introductions, and sporadic community transmission disproportionately affected mobile, working-age adults. This pattern is consistent with occupational exposure, international movement among permitted travellers, and testing uptake. Compared globally, many regions experienced Delta-driven surges with broader age involvement [Reference Luo33, Reference Earnest34]; Hong Kong’s persistence of low community transmission kept overall incidence lower, with cases concentrated where importation risk and contact networks were densest. Population-normalized incidence remained low but began to show early signs of shifting risk with the introduction of Omicron variants.
January–June 2022
Semiannual incidence peaked at 27.46 per 100 000 (2017 cases among 7.35 million), with a modest male excess (28.17 vs. 26.86 per 100 000). Age-specific incidence concentrated in working-age adults and extended into younger strata, consistent with rapid household and community transmission under BA.2.2 dominance and resumption of in-person activities. The highest rates occurred in the 20–29 age group (50.74 per 100 000), followed by the 30–39 (41.56) and 40–49 (34.22) age groups, with measurable involvement in the 10–19 (25.41) and 0–9 (24.38) age groups. Despite broad spread, severe cases were concentrated in older adults, reflecting immunological vulnerability and initially lagging booster coverage.
July–December 2022
Semiannual incidence fell to 11.50 per 100 000 (859 cases among 7.47 million), with males modestly higher than females (12.08 vs. 11.00 per 100 000). The relative burden shifted towards older adults: 80+ reached 29.99 per 100 000 and 70–79 reached 14.71, while working-age adults remained elevated (20–29 at 14.24; 30–39 at 13.97). Paediatric strata declined (0–9 at 11.70; 10–19 at 6.62). This pattern is consistent with post-surge transmission on a background of heterogeneous booster uptake and partial waning, alongside phased relaxation of public health measures.
January–June 2023
Semiannual incidence was 11.08 per 100 000 (835 cases among 7.54 million), but the age distribution became strongly senior-skewed. Incidence in 80+ rose to 115.84 per 100 000, with 70–79 at 27.84 and 60–69 at 7.74; most younger strata were substantially lower. Males exceeded females (13.54 vs. 9.02 per 100 000). This pattern indicates persistent vulnerability among the oldest adults, likely due to immune waning, exposure in clinical and care settings, and seasonal respiratory mixing, despite high population-level hybrid immunity.
July–December 2023
Semiannual incidence reached 2.17 per 100 000 (163 cases among 7.53 million), with a continued male excess (2.66 vs. 1.75 per 100 000). Absolute incidence was low across all strata, yet relative burden remained senior-focused: 80+ at 23.49 per 100 000, 70–79 at 5.23, and 60–69 at 1.30; most other strata were near or below 0.5 per 100 000. This stable endemic phase reflects high hybrid immunity and mature clinical pathways.
SARS-CoV-2 variants by sex
We evaluated variant-specific prevalence among sequenced SARS-CoV-2 cases by sex to align with semiannual age-sex incidence patterns. Among 6052 sequenced cases, variants included Alpha (n = 67), Beta (n = 47), Delta (n = 577), Kappa (n = 11), Omicron (n = 3 804), recombinant forms (n = 55), and earlier unclassified variants (n = 1 488). The distribution of variants showed significant differences by sex (chi-squared 43.026; df 5; p = 3.65 × 10^-8; Figure 5a,b), reflecting the modest male excess observed in semiannual denominators during sustained transmission periods, particularly in late 2020, late 2021, and throughout the Omicron era.
Heatmap of correlations between sex, age, and SARS-CoV-2 variants. This figure presents a heatmap illustrating correlations among sex, age, and SARS-CoV-2 variants: (a) Sex and WHO-classified variants, (b) Sex and viral lineages, (c) Age groups and WHO-classified variants, and (d) Age groups and viral lineages. The colour scheme diverges, with red indicating positive correlations and blue indicating negative correlations, based on normalized Pearson correlation residuals.
Figure 5 Long description
Heatmap of Correlations Between Sex, Age, and SARS-CoV-2 Variants. This figure presents a heatmap illustrating correlations among sex, age, and SARS-CoV-2 variants: A) Sex and WHO-classified variants, B) Sex and viral lineages, C) Age groups and WHO-classified variants, and D) Age groups and viral lineages. The color scheme diverges, with red indicating positive correlations and blue indicating negative correlations, based on normalized Pearson correlation residuals.
From May to December 2021, Delta predominance coincided with higher male incidence. In contrast, the Omicron surge beginning in early 2022 was characterized by intense community spread among working-age adults and measurable paediatric involvement, also showing a modest male excess in detections.
Fine-grained lineage analysis revealed significant sex-lineage heterogeneity for Omicron and Delta (chi-squared 208.96; df 60; p = 2.20 × 10^-16). Among Omicron variants, BA.1.1, BA.1.17.2, and BF.7.14 were more prevalent among men, while BA.2.3 and BA.5.6 were more prevalent among women. For Delta, AY.122, AY.4, and AY.43 were more frequent in men, whereas AY.23 was more prevalent in women. These patterns align with the semiannual epidemiology and international evidence of male-biased risks for severe outcomes, as well as, in some waves, higher detected incidence [35–Reference Mahmoud38].
The observed sex differences likely stem from structural and behavioural determinants, including occupation, mobility, household contact structures, and healthcare-seeking and testing practices. While biological sex differences in immune response may influence outcome severity, our data do not isolate mutation-specific mechanisms linking variants to sex-differential infectivity.
SARS-CoV-2 variants by age groups
We examined variant-specific prevalence across age groups to align with the semiannual age-sex incidence patterns and the temporal dominance of circulating lineages. Using WHO-aligned age bands for Delta and Omicron, we observed a significant association between age and variant (chi-squared 129.81; df 8; p = 2.20 × 10^-16): infections among 18–29-year-olds were more frequently Delta, consistent with late-2021 transmission concentrated in mobile working-age adults, whereas infections among those aged 85 and older were disproportionately Omicron, reflecting the Omicron-era shift towards a senior-concentrated burden amid waning immunity and clinical–interface exposure. To incorporate Alpha and Beta, we harmonized age into four strata (0–17, 18–39, 40–64, and ≥ 65 years; Figure 5c,d), which demonstrated that Alpha and Delta were relatively enriched among younger populations (0–39 years), while Beta and Omicron, including recombinant forms, were more prevalent in 18–39-year-olds and, notably for Omicron, in those aged 65 and older. These age-variant patterns are temporally consistent with the semiannual epidemiology: pre-Omicron circulation and Delta predominance were associated with higher incidence among adolescents and working-age adults, while Omicron’s endemic transition coincided with a pronounced upward age shift towards older adults.
Lineage-resolved analyses further clarified age-structured dynamics within Omicron (Figure 5c,d). BA.2.2 was most prevalent among 0–17-year-olds, consistent with intense household and school-associated transmission during phased reopenings and high contact rates in children and adolescents. BA.1.1 was comparatively enriched in 18–39-year-olds, plausibly reflecting greater mobility, occupational mixing, and social activity early in the Omicron period. Later sub-lineages displayed distinct distributions: BA.5.2 exhibited a bimodal pattern, with higher prevalence in 5–17-year-olds and adults aged 65 years and older, suggesting concurrent transmission in school networks and clinical care settings against a backdrop of heterogeneous booster timing and waning. In contrast, BQ.1.1 clustered more in 18–49-year-olds and those aged 75 years and older, highlighting concurrent susceptibility in middle-aged adults (likely via workplace and community exposure) and in the oldest adults (through clinical and long-term care interfaces). Collectively, these findings integrate with the semiannual denominators: early variants and Delta tracked with youthful and working-age contact structures during periods of broader community circulation, whereas Omicron-era lineages increasingly reflected the re-stratification of risk towards older adults as overall community incidence declined. Methodologically, the coherence between age–variant associations and the semiannual incidence context emphasizes the importance of denominator-aware genomic inference and age-structured sampling to ensure representativeness as transmission, immunity, and policy environments evolve.
Potential impact of SARS-CoV-2 mutations on host cell entry
To investigate the receptor engagement potential of SRBDs across SARS‑CoV‑2 variants, we conducted in silico docking against ACE2 and statistically compared the resulting docking scores (Figure 6; Supplementary Tables S5 and S6). The analysed variants included Alpha (B.1.1.7), Delta (AY.127 and B.1.617.2), multiple Omicron lineages (BA.1, BA.2, BA.4.1, BA.4/5, BQ.1.1, CH.1.1, CM.12, EG.5.1, HK.3, HV.1, JN.1, and XBB.1.5), and the ancestral Wuhan Hu‑1 strain. The comparison indicated differences in molecular docking scores among variants (F-statistic = 4.404, p-value = 9.12 × 10⁻⁷). We treat these results as evidence of variation in predicted ACE2 engagement propensities across variants, rather than as direct measures of experimental binding.
Mean binding affinities of the SRBD of predominant SARS-CoV-2 variants to the ACE2 receptor. This bar plot presents the mean binding affinities and standard deviation of the SRBD of various SARS-CoV-2 variants to the ACE2 receptor. Notably, the Omicron JN.1 variant exhibits the strongest binding affinity (−641.57 ± 5.00), followed by Delta AY.127 (−618.78 ± 11.65), Delta B.1.617.2 (−611.25 ± 11.51), and Omicron BQ.1.1 (−610.74 ± 5.51), all surpassing the original Wuhan Hu-1 variant (−568.71 ± 4.70) in binding strength.
Figure 6 Long description
Mean Binding Affinities of the SRBD of Predominant SARS-CoV-2 Variants to the ACE2 Receptor. This bar plot presents the mean binding affinities and standard deviation of the SRBD of various SARS-CoV-2 variants to the ACE2 receptor. Notably, the Omicron JN.1 variant exhibits the strongest binding affinity (−641.57 ± 5.00), followed by Delta AY.127 (−618.78 ± 11.65), Delta B.1.617.2 (−611.25 ± 11.51), and Omicron BQ.1.1 (−610.74 ± 5.51), all surpassing the original Wuhan Hu-1 variant (−568.71 ± 4.70) in binding strength.
Subsequent pairwise comparisons using Tukey’s HSD identified differences in docking scores between specific variant pairs. In these comparative analyses, the JN.1 variant consistently exhibited a more negative docking score than many other variants, indicating stronger predicted ACE2 engagement within this in silico simulation framework. We emphasize that docking predictions are proxies under model assumptions and should be interpreted comparatively; they are not stand-alone evidence of transmissibility or absolute affinity.
Prior structural work has shown that substitutions such as N501Y, K417N, and E484K can modulate RBD–ACE2 interactions by altering hydrogen bonding and van der Waals contacts at the interface [Reference Huang39, Reference Cueno and Imai40]. JN.1 carries a distinct constellation of substitutions within the spike protein, including I332V, K356T, R403K, V445H, L452W, and L455S, which may contribute to its more favourable docking scores by reshaping local interactions or the RBD’s conformational preferences. We note, however, that docking does not fully capture spike conformational dynamics, glycosylation, glycan shielding, or epistatic effects among mutations.
Other variants, such as B.1.1.7, CM.12, EG.5.1, HK.3, HV.1, and XBB.1.5, showed less favourable docking results compared with JN.1 in these simulations. Differences across variants are consistent with the presence or absence of mutation constellations that are known to influence ACE2 engagement and/or immune escape. For example, B.1.1.7 carries N501Y but lacks several additional substitutions present in JN.1 that may further enhance modelled ACE2 contacts.
Quantitatively, the Wuhan Hu‑1 reference strain showed a mean docking score of −568.71 (95% CI: −577.91 to −559.51). Delta variants B.1.617.2 and AY.127 exhibited more negative means of −611.25 (95% CI: −633.82 to −588.69) and − 618.78 (95% CI: −641.61 to −595.95), respectively. Omicron BA lineages had mean docking scores ranging from −574.86 (95% CI: −585.32 to −564.39) to −604.68 (95% CI: −645.61 to −563.75). For epidemiological context, reported basic reproduction numbers were approximately 1.20 for Wuhan Hu‑1 in Hong Kong [Reference Yang6] and 2.20–2.96 in Mainland China [Reference Li41, Reference Deng42]; 3.00–3.60 for Delta [Reference Yang6, Reference Luo33, Reference Zhang43]; and ≥ 9.3 for Omicron BA lineages in Hong Kong [Reference Yang6]. These comparisons emphasize that while docking scores infer the favourability of spike-ACE2 interactions, docking alone does not translate to transmission advantages at the population level: Omicron BA variants achieved high growth despite mixed or modest shifts in predicted ACE2 engagement, reflecting the important roles of immune escape, entry pathway usage, replication kinetics, and the prevailing immunity landscape.
Overall, our docking analyses reveal statistically detectable differences in modelled RBD-ACE2 engagement across variants and identify JN.1 as comparatively favourable within this framework. We interpret these results as hypothesis-generating indicators of relative ACE2 engagement, helping to prioritize variants for deeper structural interrogation. We do not infer causality between docking scores and transmissibility; instead, we contextualize docking-derived rankings against independent epidemiological metrics and acknowledge that multiple viral and host factors jointly determine real-world transmission.
Discussion
Evolution of SARS-CoV-2 variants in Hong Kong
This study provides a comprehensive overview of the evolution of SARS-CoV-2 in Hong Kong over a four-year period. Following the confirmation of the first COVID-19 case on 23 January 2020, initial control measures were swiftly implemented, including border restrictions from Mainland China and the adoption of work-from-home arrangements. By early March 2020, an increase in travel-related cases, coinciding with global outbreaks, led to the identification of 15 new circulating lineages in Hong Kong and prompted the closure of borders to non-residents. The first two waves of the pandemic, each lasting approximately 2 months, revealed 18 distinct lineages from 318 genomes, with the B, B.1, and B.3 variants predominating.
While these early pandemic responses effectively curtailed viral spread, subsequent waves of infections persisted despite stringent control measures. The third and fourth waves, spanning from July to September 2020 and late November 2020 to April 2021, respectively, were characterized by the emergence of mutations such as D614G and P314L, which indicated potential for future waves due to their enhanced transmissibility and virulence [Reference Hou44, Reference Plante45]. Recent research has linked these genetic variations, along with the mutations R203K and G204R, to increased rates of infection, hospitalization, and mortality [Reference Anand, Long and Chandran46].
In contrast, the transmission dynamics during the fourth wave in Hong Kong diverged from global trends. The B.1.36 variant, present since the early stages of the second wave, dominated, accounting for 78% (1 029 out of 1 316) of sequenced COVID-19 cases, while the highly transmissible Alpha (B.1.1.7) and Beta (B.1.351) variants, prevalent in other countries, [Reference Cahyani30–Reference Walensky, Walke and Fauci32] accounted for only about 5%. This disparity suggests a reduction in travel-related cases and a shift towards community-acquired transmission, likely occurring among individuals in densely populated and socioeconomically disadvantaged areas [Reference Yang47].
The COVID-19 vaccination campaign in Hong Kong commenced on 26 February 2021 [48], amid the ongoing fourth wave of infections. As the vaccination programme progressed, the Delta variant (B.1.617.2 + AY*) began spreading in late March 2021 and was designated a variant of concern by June 2021 [Reference Earnest34], constituting 2.66% (35 out of 1 316) of sequenced cases during the fourth wave. In response, Hong Kong implemented various measures, including the introduction of vaccine passports in July 2021, to encourage vaccination uptake.
In September 2021, expanding vaccination eligibility to individuals aged 12 and above bolstered immunization efforts. However, the emergence of the Omicron variant, particularly the BA.1 lineage, in November 2021 posed new challenges due to its rapid spread and potential for immune evasion. Consequently, stringent social distancing measures and travel restrictions were reintroduced in December 2021 to contain the Omicron variant and prevent strain on the healthcare system.
The fifth wave of COVID-19, driven by the BA.1 and BA.2 lineages, coincided with the phased reopening of schools [Reference Tsang49], emphasizing the need to protect younger populations while ensuring educational continuity. This led to the extension of vaccination eligibility to children aged 5–11 in February 2022, along with the introduction of booster doses in March 2022 to enhance immunity against emerging variants [50]. By April 2022, the fifth wave had subsided, allowing for the gradual easing of social distancing measures and a cautious return to normalcy.
The sixth wave, commencing in May 2022 and peaking between June and July 2022, was primarily driven by the BA.2.2 variant and a significant increase in BA.5 sub-lineages. This wave underscored the dynamic evolution of SARS-CoV-2, with the emergence and rapid spread of these variants potentially impacting vaccine effectiveness and public health strategies. The slight increase in mutation rates during this period (54.51 ± 2.75 mean substitution frequency, compared to 52.00 ± 3.09 during the fifth wave) indicates ongoing viral evolution, contributing to changes in transmissibility, immune evasion, and disease severity (Figure 2a,c).
The variance in lineage distribution between Hong Kong and Guangdong following the reopening of borders suggests that the predominant SARS-CoV-2 variants in Hong Kong did not originate from Mainland China but rather from alternative sources or local transmission. In Guangdong, the prevalent DY variants differ from those in Hong Kong, where BA.2.2 and BA.5 sub-lineages dominate. This indicates that China’s Digital Health Certificate has been effective in preventing the spread of the virus across reopened borders. It also emphasizes the critical need for continuous genomic surveillance and inter-regional cooperation to monitor and address SARS-CoV-2 transmission. The distinct lineage profiles observed may reflect variations in public health interventions, population immunity, and travel behaviours, necessitating tailored approaches to pandemic management. These findings highlight the importance of adaptive, evidence-based public health responses to effectively address the evolving challenges posed by SARS-CoV-2.
As 2023 progressed, the relative stability in lineage distribution observed in the second and third quarters, with the continued prevalence of BA.2.3.20 and the emergence of new lineages such as CH.1.1 and EG.5, indicates a complex interplay between viral evolution and public health measures. The persistence of the XBB.1.5 lineage at lower frequencies and the gradual increase of CH.1.1 towards the end of the year suggest that while no new dominant lineage emerged, the viral population remained diverse. This stabilization in the viral population, with no dramatic increases in any single lineage, reflects the effectiveness of ongoing public health measures and vaccination campaigns in controlling the spread of new variants. However, the detection of new lineages, even at low levels, underscores the need for continuous monitoring and research into the implications of these variants on disease transmission and severity.
While we observed trends that may reflect the influence of public health interventions and shifts in immunity, we were unable to explore factors such as vaccination status, prior infection, or clinical outcomes in greater detail because those data were not consistently available at the individual level. We acknowledge that these are important variables that likely shaped variant dynamics. Given these limitations, we focused our analysis on the patterns that could be drawn from the available genomic and demographic data. We hope future work with more integrated datasets can build on this and provide a deeper understanding of how these factors interact.
Sex and age susceptibility to SARS-CoV-2 variants in Hong Kong
Understanding the age distribution of cases is crucial for public health planning. It helps identify which age groups may require more targeted interventions, informs resource allocation, and supports the design of age-appropriate health messaging. For example, the high proportion of cases among working-age adults may point to occupational or social exposure, while the substantial share among the elderly underscores the need to continue protecting vulnerable populations.
These patterns across age groups and sex reflect a combination of biological, social, and behavioural factors. For instance, the higher proportion of female cases in certain adult age groups may be related to occupational exposures (such as healthcare and caregiving roles), while the male predominance in older age groups may reflect differences in susceptibility, comorbidities, or health-seeking behaviour.
The observed sex-specific associations with different SARS-CoV-2 variants may be influenced by biological, behavioural, and social factors. The differential likelihood of contracting specific SARS-CoV-2 lineages by sex may stem from hormonal influences, including the immunomodulatory effects of oestrogen and progesterone [Reference Jones51, Reference Robinson52]. Genetic factors, including sex-linked variations, may also impact immune responses [Reference Berghöfer53, Reference Griesbeck54], while behavioural differences, such as variations in social interactions, occupational exposures, and adherence to public health measures, could contribute to the observed patterns. Future investigations are necessary to explore these hypotheses and unravel the intricate interplay of factors influencing sex-specific susceptibility to SARS-CoV-2 variants.
Age-related susceptibility to SARS-CoV-2 variants extends beyond biological factors and is also influenced by the virus’s spatiotemporal distribution. Positive associations of the Delta and Omicron variants with the 18–29 and 85 years and older age groups, respectively, were noted, correlating with the transmission peaks of these variants across pandemic waves. The predominance of the Delta variant in Hong Kong in the latter half of 2021 coincided with increased social activity among younger adults, who were less likely to be fully vaccinated compared to older individuals. The increased social interactions and lower vaccination rates among the 18–29 age group likely facilitated the transmission of the Delta variant within this demographic. In contrast, the emergence of the Omicron variant towards the end of 2021 and into 2022 was associated with a higher prevalence among the elderly population (85 years and older). This can be partly attributed to waning immunity in older adults, who were among the first to receive vaccinations, combined with the Omicron variant’s heightened transmissibility and immune-evasion characteristics. Additionally, the elderly are more susceptible to severe COVID-19 outcomes, leading to increased detection rates and more frequent testing within this age group during the Omicron wave.
The underlying mechanisms behind the sex- and age-specificity of certain SARS-CoV-2 variants remain unclear. Understanding these differences in susceptibility can enhance medical interventions and identify high-risk populations. The complexities of patient susceptibility, morbidity, and mortality to COVID-19 stem from a multitude of factors. Therefore, we recommend future investigations and comprehensive sampling efforts to gather pertinent biological data (e.g., gene and protein expression, hormone levels, transcription factor activation, post-translational modifications, and comorbidities) as well as socio-economic data (e.g., occupation, education level, household income) over an extended period, alongside viral genome sequencing.
Although we acknowledge that the available metadata on age and sex was incomplete and could introduce some degree of selection bias, we minimized this risk by applying consistent inclusion criteria across all time periods. Specifically, we retained only high-quality human-origin sequences with sufficient genome length, coverage, and metadata completeness. No manual filtering was applied, and all eligible sequences were included in the analysis. With this approach, we were able to reduce the likelihood of sampling-related bias in the observed demographic trends. Notably, future studies with more complete metadata would enable multivariable analyses that simultaneously adjust for sex and age, which could yield deeper insights into demographic susceptibilities.
SRBD-ACE2 binding across SARS-CoV-2 variants
This study also evaluated the receptor engagement potential of SARS‑CoV‑2 variants by performing in silico docking of the spike RBD to ACE2, with particular focus on Omicron sub-lineages circulating in Hong Kong, including JN.1. In these simulations, several variants, most notably JN.1, exhibited more favourable (more negative) docking scores compared with many earlier lineages, indicating comparatively stronger predicted ACE2 engagement within the constraints of the docking framework. We interpret these docking outputs as comparative, hypothesis‑generating indicators of RBD–ACE2 interaction strength rather than direct measurements of biophysical affinity or transmissibility.
The residue‑level context aligns with prior structural and biochemical investigations. Earlier studies have shown that substitutions such as N501Y, K417N, and E484K can modulate RBD–ACE2 interactions and antigenicity at the interface by altering hydrogen bonding and van der Waals contacts [Reference Huang39, Reference Cueno and Imai40]. Specifically, N501Y has been reported to increase RBD–ACE2 binding by introducing additional hydrogen bonds and improving interface packing [Reference Tian55], whereas changes at K417 and E484 affect ACE2 interactions and are associated with reduced antibody neutralization, contributing to immune escape [Reference Jangra56]. In our analyses, JN.1 carries a distinct constellation of substitutions within the SRBD, including residues shared with other Omicron lineages (e.g., G339H, S371F, S373P, S375F, D405N, R408S) as well as unique changes such as I332V, K356T, R403K, V445H, L452W, and L455S not present in earlier variants. Among these, L452W and L455S plausibly introduce stabilizing interactions and could support more favourable docking poses. Together, the combination of pre‑existing key residues and JN.1‑specific substitutions provides a mechanistic rationale for its comparatively favourable docking scores.
At the same time, we do not directly equate docking‑derived results with transmissibility. Population‑level transmission reflects a composite of factors beyond RBD–ACE2 interactions: spike conformational dynamics (RBD up/down), S1/S2 cleavage and fusion activation, glycosylation and glycan shielding, replication efficiency, and, during the Omicron period in Hong Kong, substantial immune escape under high population immunity. Consistent with this, our epidemiological context shows that Omicron BA lineages achieved markedly higher reproduction numbers in Hong Kong (≥9.3) than Delta (3.0–3.6) or Wuhan Hu‑1 (approximately 1.20 in Hong Kong and 2.20–2.96 in mainland China) [Reference Yang6, Reference Luo33, Reference Li41–Reference Zhang43], despite mixed or modest changes in predicted ACE2 engagement. These observations highlight that docking should be integrated with epidemiological indicators and experimental measurements (e.g., surface plasmon resonance, biolayer interferometry, or cell-based infectivity assays) rather than interpreted as stand‑alone evidence of transmissibility.
This study has several limitations. First, it relies on SARS-CoV-2 genomes deposited in NCBI and GISAID. Although these repositories enable territory-level, time-resolved analyses, sampling is non-random, and submission intensity varies across waves and sites, with incomplete metadata for key covariates (location, clinical setting/severity, and vaccination). This limits fine-grained representativeness and precludes formal spatial inference. While our coverage tracked epidemic magnitude and lineage turnover mirrored territory-wide surveillance, residual reporting and clinical-pathway biases may persist. Therefore, we restrict claims to citywide comparative and temporal patterns, reported lineage turnover across pandemic waves and age strata, and we recommend future integration of enriched metadata to enable reweighting and mixed-effects models. Second, our inclusion criteria (genome length ≥ 29 000 bp, coverage ≥90%, ambiguous bases <5%, complete collection dates) prioritize analytical validity but may exclude some valid partial genomes and introduce bias towards higher quality, possibly more recent samples. Third, primary outputs (lineage proportions and mutation prevalence) are descriptive surveillance measures without multiple hypothesis testing. Where future work entails families of statistical tests, we recommend controlling the false discovery rate and clearly distinguishing exploratory from confirmatory analyses. Fourth, demographic metadata were incomplete (6 052 with sex, 4 671 with age), introducing potential selection bias in age- and sex-stratified summaries. Finally, our docking analyses provide comparative, hypothesis-generating indicators of modelled RBD–ACE2 engagement rather than definitive biophysical measurements. They do not fully capture protein dynamics, glycosylation, post-translational modifications, membrane context, or cell-specific factors, and thus cannot establish causality with transmissibility or severity. Orthogonal validation (SPR/BLI for KD, pseudovirus/live-virus assays, and glycan-aware MD/free-energy methods) is warranted to corroborate and refine these in silico findings.
Conclusion
This study provides a detailed examination of the evolution of SARS-CoV-2 variants in Hong Kong over a four-year period, highlighting the complex interplay between viral mutations, public health responses, and demographic factors. The findings emphasize the virus’s dynamic nature, revealing how specific variants have emerged, spread, and adapted in response to changing social behaviours and vaccination strategies. The distinct lineage profiles observed in Hong Kong, particularly compared with neighbouring Guangdong, emphasize the importance of localized public health measures and the need for ongoing genomic surveillance to monitor the evolving SARS-CoV-2 landscape.
Critical steps must be taken to enhance our response to COVID-19 and future infectious disease threats. First, by enhancing our genomic surveillance capabilities, we can detect and respond swiftly to new variants and pathogens before they escalate into larger outbreaks. This proactive approach is vital in identifying potential zoonotic diseases, those that can jump from animals to humans, such as avian influenza, Ebola, and other coronaviruses. Strengthening vaccination campaigns, particularly for high-risk populations, will be crucial in mitigating the impact of future outbreaks, as demonstrated by the rapid development and deployment of COVID-19 vaccines.
Moreover, the lessons learned from the COVID-19 pandemic should inform our broader approach to global health preparedness. This includes fostering international collaboration for data sharing, research, and resource allocation, ensuring that we can respond effectively to emerging threats. Public health systems must remain adaptable and ready to implement flexible strategies tailored to the unique characteristics of new infectious agents.
Finally, investing in public health infrastructure, education, and community engagement will enhance our resilience against future pandemics. By understanding the socio-economic and environmental factors that contribute to the emergence of infectious diseases, we can develop comprehensive strategies that address the root causes of outbreaks.
In essence, while the fight against SARS-CoV-2 continues, we must also prepare for the next potential pandemic. This study illustrates how integrating genomic surveillance, demographic analysis, and molecular modelling can provide a clearer picture of how variants evolve, spread, and affect different population groups. Such an approach can support more targeted and responsive public health planning, particularly in urban settings like Hong Kong. The insights gained from this work contribute to the broader effort of building a stronger, more adaptive surveillance framework, one that can help manage current risks and better prepare us for emerging outbreaks. The future of public health relies on our collective ability to learn, adapt, and collaborate in the face of evolving threats.
Supplementary material
The supplementary material for this article can be found at http://doi.org/10.1017/S0950268826102040.
Data availability statement
All 22 340 viral genome sequences analysed in this study were derived from the comprehensive repositories of approximately 8.5 million and 16.4 million SARS-CoV-2 sequences available in NCBI and GISAID, respectively, as of January 2024. The dataset utilized in this research is accessible through GISAID at gisaid.org using the identifier EPI_SET_250105dz or via the DOI 10.55876/gis8.250105dz, in accordance with GISAID’s data access and sharing policies.
Acknowledgements
We sincerely acknowledge all NCBI and GISAID data contributors, particularly the authors and originating laboratories of the sequences and metadata that informed these findings.
Author contribution
All the authors included in this study had substantial contributions to this research article and have read and approved the final version of this paper. Conceptualization: J.D.P., K.L.C., and K.L.Y.; Data curation: J.D.P., M.J.P., and K.L.Y.; Formal analysis: J.D.P., M.A.M., R.R., M.J.P., and K.L.Y.; Supervision: K.L.Y.; Validation: M.A.M. and R.R.; Visualization: J.D.P. and K.L.Y.; Data interpretation: J.D.P., M.A.M., R.R., M.J.P., K.L.C., and K.L.Y.; Writing – original draft: J.D.P., M.A.M., and K.L.Y.; Writing – review and editing: J.D.P., R.R., K.L.C., and K.L.Y.
Funding statement
We also thank the HKUST-CIL Joint Laboratory on Advanced Environmental Health Technologies for their financial support.
Competing interests
The authors have no competing interests to disclose.
Open access
