Biology Reference
In-Depth Information
the population. The algorithm now enters the next iteration, and, again, it
will evaluate the fitness of the new population and so forth until it reaches
stop criteria—that is, either a new population does not have better fitness
than the previous one, the objective function is satisfied, or it has reached
the maximum number of population generated. In general, GA performs
well especially in a complex search spaces, although the solution may only
be local optima.
The most common application of GAs in epidemic modeling is through
the use of the Genetic Algorithm for Rule Prediction (GARP) (Stockwell and
Peters 1999). This is a method widely used in the biodiversity discipline, and
it projects the potential distribution of animals and plants in unknown areas
based on a set of if-then rules relating point-occurrence data and associ-
ated geographical and environmental variables. The GA is used to develop
or search the rules that ultimately give the best description of the species
distribution based on ecological and environmental factors. Consequently,
with the if-then constraints, one can determine the species distribution in
an unknown area and possibly at different times, given particular ecological
and environmental information. In infectious disease transmission studies,
GARP can be used to predict the geographical distribution of disease vectors
in indirectly transmitted diseases (this approach is also known as Ecological
Niche Modeling [ENM]). Levine et al. (2004), for example, has successfully
used GARP to predict the geographical location of
Anopheles gambiae
com-
plexes in Africa using 12 environmental layers of the available 14. Beard et al.
(2003) used GARP to predict the distribution of the vector for Chagas dis-
ease,
Triatoma gerstaeckeri
, in southern Texas. In addition to predicting vec-
tor distribution, GARP can also be employed to project disease incidence as
demonstrated by Peterson et al
.
(2004) in their filovirus disease study. In that
case, GARP was appropriate to use since neither the filovirus reservoir nor
the transmission mechanism was known.
3.5.1.5 Ecological Niche Modeling (ENM)
Ecological Niche Modeling (ENM) relates ecological niche with a species'
geographical distribution. Ecological niche describes a set of conditions
needed by a species to maintain its population without any immigration
(Peterson 2006). In ENM, recorded existence of a species in any area is char-
acterized by its environmental factors. Based on this collective information,
ENM predicts the spatial distribution of the said species in an unsampled
area or landscape. In terms of infectious disease modeling, ENM is suitable
for predicting the geographical distribution of the vector. However, it should
be noted that this is not exactly equivalent to predicting the risk and preva-
lence of a disease. Occurrence of vector does not directly relate to the prob-
ability of disease transmission, and, therefore, further modeling is needed to
relate a vector's distribution with transmission probability. Refer to Soberon
and Peterson (2005) for further details on ENM.
