Biology Reference
In-Depth Information
i)
Genotype
, where full genotypes were reported; ii)
Phenotype
, where full
serological diagnoses (anti-Fy
a
and anti-Fy
b
sera) were used; iii)
Promoter
,
if results only stated antigen expression or non-expression (no distinc-
tion between Fy
a
and Fy
b
); iv)
Phenotype-a
, if only Fy
a
was tested for (no
distinction between Fy
b
and the negative phenotype); and v)
Phenotype-b
,
if only Fy
b
was tested for (no distinction between Fy
a
and the negative
phenotype).
This dataset formed the evidence-base for a geostatistical model used to
predict a continuous map of the prevalence of the Duffy-negative phenotype
[Fy(a−b−)], the phenotype encoded by the
FY
*
B
ES
/
FY
*
B
ES
genotype. The
FY
*
A
ES
is highly infrequent and so was modelled as a small constant rather
than a spatially variable allele. Although not all of the five data types directly
informed the prevalence of Duffy negativity, the model was able to infer
useful information directly from each (by ruling out certain genotypes),
so all were included as model inputs. To allow this, the model structure
was based on the frequencies of the two loci involved in the system: the
promoter type (determining expression
versus
non-expression at the T-33C
site) and the coding region (determining the G125A polymorphism). The
Bayesian framework considered the two loci as spatially independent ran-
dom fields, and used the survey data to map the frequency of the expression
of each variant. Each of the data types informed different aspects of these
loci: some informed both loci and others excluded possible variants. The
model also incorporated a land cover variable to distinguish sub-Saharan
African populations from others on the continent. This was to inform the
model of the high probability of association between silencing mutation of
the Fy
b
variant and the
FY
*
B
ES
allele, which was observed in high frequen-
cies across sub-Saharan Africa.
The Bayesian model-based geostatistical framework predicted Duffy
group expression frequencies in all geographic locations across a 5 × 5 km
grid to generate a continuous global surface of each variant. The Duffy neg-
ativity phenotype was expressed by the squared frequency of the
FY
*
B
ES
allele. The map reveals that the homozygous null phenotype is highly con-
strained to sub-Saharan African populations, with localized high-frequency
areas in the Americas. Historical perceptions have supported the assump-
tion that
P. vivax
is absent from much of the African continent (
Rosenberg,
2007
). However, evidence of autochthonous transmission within Africa
indicated that areas of the continent should not be excluded a priori. This
map, therefore, provided an evidence-based exclusion layer for Duffy-nega-
tive populations resistant to
P. vivax
infection in Africa.
