Forensic interpretation of molecular variation on networks of disease transmission and genetic inheritance

Abstract

This paper describes the inference-on-networks (ION) framework for forensic interpretat ION of molecular typing data in cases involving allegations of infectious microbial transmission, association of disease outbreaks with alleged sources, and identifying familial relationships using mitochondrial or Y chromosomal DNA. The framework is applicable to molecular typing data obtained using any technique, including those based on electrophoretic separations. A key insight is that the networks associated with disease transmission or DNA inheritance can be used to define specific testable relationships and avoid the ambiguity and subjectivity associated with the criteria used for inferring genetic relatedness now in use. We discuss specific applications of the framework to the 2003 severe acute respiratory syndrome (SARS) outbreak in Singapore and the 2001 foot-and-mouth disease virus (FMDV) outbreak in Great Britain.

Keywords: Forensic; Genetics; Inference; Networks; Transmission.

Figure 1

A notional transmission tree. Each node (dot) represents an individual, and S₁ and S₂ are genetic sequences obtained from isolates that come from two nodes of interest (e.g. two infected persons), marked in blue and red, respectively. H₀, H₁, and H_j represent different hypotheses about the source of the genetic material found in the red node. Asterisks mark two nodes separated by seven transmission steps on this tree.

Figure 2

Calculations of P(M ≤ 5|k) for networks with successively larger numbers of generations G. This shows that as long as G > 2M, the precise number of generations used to determine P(M) is not critical.

Figure 3

Estimated posterior probability that a pair of SARS isolates arose from direct transmission given that their sequences differ by k substitutions. Solid lines—using the indicated empirical or model P(M) distribution and the Poisson distribution for P(k|M) with γ = 3; Broken lines—calculated with γ = 1 and γ = 5, respectively, with the P(M) for outbreak TSSH1.

Figure 4

Predicted posterior distribution P(M = 1|k) based on data from 25. Gray curves are based on 20 randomly generated transmission trees; Black points are based on the tree published in 25.