Identifying likely transmission pairs with pathogen sequence data using Kolmogorov Forward Equations; an application toM.bovisin cattle and badgers

Abstract

18Established methods for whole-genome-sequencing (WGS) technology allow for the detection of 19 single-nucleotide polymorphisms (SNPs) in the pathogen genomes sourced from host samples. The 20 information obtained can be used to track the pathogen's evolution in time and potentially identify 21 'who-infected-whom' with unprecedented accuracy. Successful methods include 'phylodynamic 22 bovine/zoonotic tuberculosis, bTB) infection. 31Our results show that, even with such limited data and low diversity, the computation of the 32 transmission probability between host pairs can help discriminate between likely and unlikely 33 infection pathways and therefore help to identify potential transmission networks, but can be 34 sensitive to assumptions about within-host evolution. 35 36 Keywords: bovine tuberculosis, inter-species transmission, contact network, within-host evolution, 37Mycobacterium bovis, Woodchester Park 38 39 40 contacts between them as links (edges). 46 An important distinction exists between the contact network and the transmission network: while 47 the former includes all potential transmission contacts, the latter is a subset of the former describing 48 pathogen transmission patterns 5,6 . Identifying the transmission network, even when the contact 49 network is well described, can be a challenging filtering process informed by multiple factors. 50Techniques are still needed to disentangle these factors, using the different sources of information 51 (evolutionary, immunological, and epidemiological) available to infer likely transmission pathways. 52Most importantly, we wish to know "how likely is it that individual A infected individual B?", or 53 "how likely is it that a third unsampled individual was involved in the transmission chain between 54 individuals A and B?", the key questions in forensic or 'precision' epidemiology. The answers to 55 these questions, and transmission pathway reconstruction, are important for gathering information 56 about outbreaks, to shed light on transmission dynamics, and to help infer epidemiological 57 parameters. 58Whole genome sequencing (WGS) can be used to detect polymorphisms in a genome with high 59 resolution, and therefore discriminate between closely related strains. Polymorphisms are caused by 60 errors that occur during pathogen replication within the host. Generally these single nucleotide 61 polymorphisms (SNPs) are considered to be neutral in bacterial species within the timescale of 62 disease outbreaks 7 . In the absence of horizontal genetic transfer, tracking these SNPs would be 63 expected to follow the pattern of transmission. In combination with an increasing ability to extract 64 genetic material (either directly from clinical samples or from cultured isolates) and with rapid and 65 minimal processing, large-scale characterization of populations of pathogen genomes is now 66 4 possible 8-10 . These advances have proven to be transformative for forensic epidemiology, 67 especially when populations can be sampled densely. 68The observed gen...

Phylodynamic analysis of an emergentMycobacterium bovisoutbreak in an area with no previously known wildlife infections

“…The pairwise transmission probability was calculated using the Kolmogorov Forward Equations (KFEs) [41], and the pairwise transmission probability matrix was reported in Figure 4. The KFEs methodology was used to calculate the probability of observing a pair of bTB infected hosts given their sampling time and the genetic divergence between their bacterial isolates, assuming that a direct transmission occurred between the pair.…”

“…Results showed that random trees had a median [95 th quantile] of 17 [13][14][15][16][17][18][19][20][21] cattle-to-cattle, 26 [21][22][23][24][25][26][27][28][29][30][31] badger-to-badger, 14 [10][11][12][13][14][15][16][17][18][19] cattle-to-badger, and 6[2-10] badger-to-cattle transmissions ( Figure S4.3). We computed the most likely transmission tree by selecting the transmissions with highest probability within the sampled animals while avoiding loops in the tree (see [41]). The best tree ( Figure 4) showed that most of the transmissions in this system likely happened within-species, i.e.…”

“…We calculated the transmission probability between pairs of animals where the M. bovis sequences have been sampled. We calculated this probability using the Kolmogorov-Forward Equations (KFE) methodology [41]. The KFEs consist of a set of ordinary differential equations which track the probability of a system to be in a given state through time [57][58][59].…”

“…For badgers we assumed a constant death rate, based on the observation that less than 0.1% of individuals would survive past 8-years of age [27]. The epidemiological parameters were chosen according the most recent literatures (see [41]), and in order to account for their variability, for each pairwise transmission we tested a 1'000 combination of randomly selected parameters combinations and chose the one returning the highest probability.…”

“…Following Rossi et al [41], we assembled the most likely transmission tree by progressively selecting the pairs with the highest transmission probability, and excluding those not possible given the previously selected ones (e.g. if A®B and B®C are selected, B®A, C®B and C®A were going to be a priori excluded).…”