key: cord-0950534-4oa0gsos
authors: Hahn, Georg; Cho, Michael H.; Weiss, Scott T.; Silverman, Edwin K.; Lange, Christoph
title: Unsupervised cluster analysis of SARS-CoV-2 genomes indicates that recent (June 2020) cases in Beijing are from a genetic subgroup that consists of mostly European and South(east) Asian samples, of which the latter are the most recent
date: 2020-06-30
journal: bioRxiv
DOI: 10.1101/2020.06.22.165936
sha: f83f6b15ad4cfc5275c974bd6736865d88b1672d
doc_id: 950534
cord_uid: 4oa0gsos

Research efforts of the ongoing SARS-CoV-2 pandemic have focused on viral genome sequence analysis to understand how the virus spread across the globe. Here, we assess three recently identified SARS-CoV-2 genomes in Beijing from June 2020 and attempt to determine the origin of these genomes, made available in the GISAID database. The database contains fully or partially sequenced SARS-CoV-2 samples from laboratories around the world. Including the three new samples and excluding samples with missing annotations, we analyzed 7, 643 SARS-CoV-2 genomes. Using principal component analysis computed on a similarity matrix that compares all pairs of the SARS-CoV-2 nucleotide sequences at all loci simultaneously, using the Jaccard index, we find that the newly discovered virus genomes from Beijing are in a genetic cluster that consists mostly of cases from Europe and South(east) Asia. The sequences of the new cases are most related to virus genomes from a small number of cases from China (March 2020), cases from Europe (February to early May 2020), and cases from South(east) Asia (May to June 2020). These findings could suggest that the original cases of this genetic cluster originated from China in March 2020 and were re-introduced to China by transmissions from samples from South(east) Asia between April and June 2020.

One important aspect being investigated about the ongoing pandemic of the SARS-CoV-2 virus pertains to its geographic progression around the world and the linked question of possible viral mutations during viral spread.

We focus on the recent report of three newly diagnosed cases in Beijing and their viral genomes that were reported in June 2020. One working hypothesis is that these new cases originated from Europe (Xiaohua, 2020; Liu and Nebehay, 2020) . We utilized an unsupervised clustering approach where results of the previously sequenced viral genomes matched the history of the geographic spread of SARS-CoV-2 (Hahn et al., 2020a) .

The approach we employ is an unsupervised clustering technique considered in Hahn et al. (2020a) . In this paper, we consider the computation of a similarity measure on the Hamming matrix of the SARS-CoV-2 genome samples available in the GISAID database (Elbe and Buckland-Merrett, 2017; Shu and McCauley, 2017) . The Hamming matrix encodes the difference of all samples compared to the reference genome, and the similarity measure being employed is the Jaccard matrix (Prokopenko et al., 2016; Schlauch et al., 2017) . Visualizing the first two principal components of the Jaccard similarity matrix revealed that the samples from North America and Europe fall into four branches. Those branches originated from the origin, which contained samples from Wuhan, other parts of China, and geographically proximate areas such as Southeast Asia, thus reflecting the geographic progression of the virus across the globe.

In this paper we repeat the PCA (principal component) analysis of the Jaccard similarity measure on the Hamming matrix of Hahn et al. (2020a) . The dataset under investigation is a recent timestamp (14 June 2020) of single stranded RNAs of SARS-CoV-2 viruses, extracted from SARS-CoV-2 patients around the world and downloaded from the GISAID database. We contrast the three newly found SARS-CoV-2 genomes in Beijing, available on GISAID under the accession numbers EPI ISL 469254, EPI ISL 469255 and EPI ISL 469256, to the other 7, 640 SARS-CoV-2 patients without missing entries available in the database. The first two samples, EPI ISL 469254 and EPI ISL 469255, are extracted from human patients, whereas EPI ISL 469256 is an environmental sample. We show that all three samples indeed fall into one of the European and South(east) Asian branches of the SARS-CoV-2 genome reported in Hahn et al. (2020a) , which also contain a small number of older cases from China stemming from March 2020. As the most recent cases in these branches are almost exclusively from South(east) Asia, this could suggest that the new cases in Beijing were re-introduced by transmissions from South(east) Asia.

We downloaded all complete nucleotide sequences (defined as having a length ≥ 29, 000) of the SARS-CoV-2 genome available from the GISAID database (Elbe and Buckland-Merrett, 2017; Shu and McCauley, 2017) with a timestamp up to 14 June 2020. This dataset contains 45, 895 different sequences.

We proceed by keeping only those samples having a valid collection timestamp, and filter for sequences without reading errors, that is without missing base pair entries. Afterwards, we group samples into 13 distinct categories using their geographic location tag. Those categories were: (1) China (all provinces apart from Wuhan), (2) Wuhan, (3) Korea, (4) Europe (all countries apart from Italy), (5) Italy, (6) Africa, (7) Australia and New Zealand, (8) South(east) Asia, (9) North America (USA, Canada), (10) Central America (including Mexico), (11) South America, (12) Japan, (13) Iran.

We employ the anchor sequences of (Hahn et al., 2020a, Section 2) to align all sequences.

Samples without a match of the anchor sequence were discarded, leaving n = 7, 643 samples for analysis including the new samples from China.

After aligning with the anchor, all samples as well as the SARS-CoV-2 reference sequence (available on GISAID under the accession number EPI ISL 412026 ) were aligned in order to establish a window in which all sequences had reads. This window comprised a length of p = 28, 784.

We are interested in measuring the similarity between the samples in order to group them.

Instead of comparing the samples directly, to have a common ground for comparison, we compare their differences to the reference sequences.

Similarly to Hahn et al. (2020a) , we denote with X ∈ S n×p the matrix of all n nucleotide sequences, where S ∈ {A, C, G, T }. We denote with r ∈ S p the reference genome from GISAID. We compute the Hamming matrix Y ∈ {0, 1} n×p encoding the differences of each nucleotide sample compared to the reference sequence after alignment. Precisely, we define Y ij = 1 if and only if X ij = r j for sample i, and Y ij = 0 otherwise. Each entry Y ij = 1 thus indicates a mismatch at position j of the sample i with the reference sequences. The row sums of Y are the Hamming distance of each sample to the reference sequence.

The resulting Hamming matrix Y is analyzed with the locStra methodology of Hahn et al. Regarding the newly found samples in Beijing, we observe that all three samples EPI ISL 469254, EPI ISL 469255, and EPI ISL 469256 remain after filtering and can be aligned to the reference sequence. (see Figure 2 ). While the majority of the European cases are from February to early May 2020, the cases from South(east) Asia belong to the most recent time period from May to June 2020. This could suggest that the new cases in Beijing were caused by transmissions that are linked to cases from South(east) Asia.

We repeated an unsupervised clustering approach of SARS-CoV-2 samples available on the GISAID database, using the Jaccard similarity index computed on the matrix of Hamming differences Here, we were particularly interested in three newly found samples with GISAID accession numbers EPI ISL 469254, EPI ISL 469255, and EPI ISL 469256 which were discovered in Beijing in June 2020. One working hypothesis is that the newly found cases were "imported" from outside.

While the new cases do not form a new cluster, they fall into distinct clusters of cases that, with the exception of a few early cases from China (March 2020), are predominantly European and South(east) Asian cases. According to the sequencing time stamps, the most recent cases in these clusters are from South(east) Asia, which could imply that the new cases in Beijing are attributable to transmissions from South(east) Asian cases.

Data, disease and diplomacy: GISAID's innovative contribution to global health

Unsupervised cluster analysis of sars-cov-2 genomes reflects its geographic progression and identifies distinct genetic subgroups of sars-cov-2 virus

locstra: Fast analysis of regional/global stratification in whole genome sequencing (wgs) studies

locStra: Fast Implementation of (Local) Population Stratification Methods (v1.3

China sees European virus strain in Beijing, WHO says more study needed

Utilizing the Jaccard index to reveal population stratification in sequencing data: a simulation study and an application to the 1000 Genomes Project

Identification of genetic outliers due to sub-structure and cryptic relationships

GISAID: Global initiative on sharing all influenza data -from vision to reality

Experts decode the genome sequence released by CDC: Beijing new virus is imported

The authors gratefully acknowledge the contributors, originating and submitting laboratories of the sequences from GISAID's EpiCoV TM Database (Elbe and Buckland-Merrett, 2017; Shu and McCauley, 2017) on which this research is based. A detailed list of contributors is available in the Supplementary Information.

Sequence data that support the findings of this study are deposited in the GISAID database with accession numbers in the range of EPI ISL 402119 to EPI ISL 469256 (https://www.gisaid.org/).

The initial methodology work for this paper was funded by Cure Alzheimer'