Improvement of Yield per se in Sugarcane - Translational Genomics for Crop Breeding: Abiotic Stress, Yield and Quality - page 226

Association Mapping breeding was highlighted early on (Jannoo et al.

1999). Only a few generations separate modern
cultivars from their interspecific founders, lim-
iting the number of meioses and consequently
the opportunity for chromosome recombination.
Moreover breeding history is characterized by
bottlenecks arising from the very limited number
of founders (Arceneaux 1967). As a result,
despite the relatively large size of the sugarcane
genome (about 10 Gb), modern germplasm is
thought to encompass a rather modest number
of LD blocks. None of the classic measures of
LD (D', r 2 ,d 2 ) exploiting allele frequency or
haplotype frequency can be calculated because
of the high polyploidy of sugarcane. However
the Fisher exact test probability can be used to
test for associations between markers. Using
information from a reference map (Hoarau et al.
2001) combined with the study of a small panel
of sugarcane cultivars, Raboin and colleagues
(2008) assessed LD in modern sugarcanes. As
predicted, LD appeared to be more extensive
than in numerous other plants, since LD drops
sharply only over a distance of 5 cM and
instances of LD blocks of 10 to 20 cM are
relatively frequent. But many LD blocks may
be missed, as the confounding effects of marker
dosage due to polyploidy are assumed to mask
many instances of linked markers (Costet et al.
2012).
Several authors have applied association map-
ping in sugarcane relative to disease and insect
resistance such as smut, African stalk borer,
pachymetra root rot, leaf scald, and Fiji leaf
gall (Raboin 2005; Wei et al. 2006; Butterfield
2007), or to yield component traits (Wei et al.
2010). These studies have led to the detection of
numerous marker-trait associations, despite the
use of a modest number of markers, far from the
number required for a “meta” genome of a culti-
var panel to be densely scanned. Sugarcane hap-
loid genome size is about 1500 cM. Therefore,
considering average LD values in sugarcane,
a minimum of 300 to 600 multi-allelic locus-
specific markers would be required to achieve a
minimum density of one or two markers every

Association studies based on linkage disequilib-
rium (LD) can also be used to tag QTLs without
necessarily calling for detailed linkage informa-
tion. Unlike linkage analysis based on controlled
progenies, LD-based studies do not require
segregating populations of known parentage.
Linkage, but also selection, or drift, in a popula-
tion are the main causes for allelic associations
to occur at a different frequency from what
would be expected if the associations were due
to random mating (Flint-Garcia et al. 2003).
LD-based studies, or association-mapping
studies, are based on existing populations or
germplasm collections, which may have major
advantages: (1) individuals in such collections
may be more or less distantly related and may
have accumulated recombination events over
many generations and consequently allow for
high-resolution mapping, (2) such collections
may already be well characterized for a range of
interesting traits, and (3) such collections may
include a large number of alleles of agronomic
value (Morgante and Salamini 2003; Rafalski
and Morgante 2004). However, association-
mapping studies suffer from certain limitations.
There is a higher probability of type I and type
II errors compared to classic bi-parental QTL
analysis (Breseghello and Sorrells 2006). Type I
errors, or detection of false marker-trait associ-
ations, may be the result of the genetic structure
within the population studied. Type II errors,
that is, the probability of missing genuine causal
associations, may result from (1) lower associa-
tions between markers and genes resulting from
the rapid decay of LD, (2) unbalanced design
resulting from the presence of alleles at distorted
frequencies, and (3) very strict genome-wise
significance thresholds resulting from the
relative independence of the many markers
tested (Carlson et al. 2004). The extent of LD
determines whether genome scans or candidate
gene association approaches can be used (Nord-
borg and Tavare 2002; Flint-Garcia et al. 2003).
The potential of LD approaches in sugarcane