Operationalizing genomic epidemiology during the Nord-Kivu Ebola outbreak, Democratic Republic of the Congo.
This repository represents the analysis and code behind Kinganda-Lusamaki et al., currently available at https://www.medrxiv.org/content/10.1101/2020.06.08.20125567v1
Analysis in the paper was performed using Jupyter notebooks and Nextstrain.
I suggest installing the Nextstrain conda environment as well as jupyter (e.g. pip install jupyterlab).
Full resolution figures can be found in ./figures.
Figure 1, "Progress of genomic surveillance over the course of the outbreak", is produced by ./notebooks/lags.ipynb.
Figure 2, "Broad scale spatiotemporal dynamics of the EVD outbreak in Nord Kivu province DRC", comprises screenshots from Auspice of the ./auspice/ebola-narrative-ms_full-build.json dataset.
Figure 3, "Transmission patterns within and between health zones", as well as corresponding supplementary figures, are produced by ./notebooks/chains.ipynb
Figure 4, "Initial genomic evidence for an infection relapse event" was created similarly to Figure 2.
Supplementary Figures 2 & 3, "Patterns of transmission between health zones (HZ)" and "Viral movements are robust to subsampling" are produced by the same notebook as Figure 3 (./notebooks/chains.ipynb).
Supplementary Figure 4, "The outbreak was sustained by movement between health zones", is produced by ./notebooks/exploded_tree.ipynb.
Supplementary Figures 5 & 6, "Genomic characterization of transmission after unsafe burial of a pastor" and "Secondary transmission associated with infection of a motorcycle taxi driver", are produced from Auspice screenshots of the ./auspice/ebola-narrative-ms_full-build.json dataset with additional manual annotations.
The sequence alignment was generated using a banded-alignment approach with manual editing in AliView. The dataset was partitioned using a script from nextstrain/ncov (python partition-sequences.py --sequences data/sequences.fasta --sequences-per-group 102 --output-dir ignore/bands/). Alignment was performed using a modified version of augur which added the --maxiterate 1000 argument to mafft (augur align --nthreads 2 --sequences results/0.fasta --reference-sequence config/reference.gb --output results/0.aligned.fasta --fill-gaps).
Each of the resulting alignments were checked by hand in AliView, with no clear misalignments identified.
Alignments were combined using a script from nextstrain/ncov (python combine-and-dedup-fastas.py --input ignore/bands/*.aligned.fasta --output data/aln.fasta).
ADAR edits, which were noticed upon manual inspection were removed using a custom script based upon Dudas et al via python scripts/remove_ADAR_edits.py --input data/aln.fasta --output results/aln.noAdar.fasta.
A number of (singlteon) SNPs flanking regions of Ns were also detected upon manual checking of the alignment which are most probably spurious sequencing errors.
These were removed via scripts/remove_SNPs_flanking_Ns.py --input aln.noAdar.fasta --output results/manual_alignment.fasta.
The final alignment consisted of 808 sequences.