Supplementary Table 9. Table showing the variance components and proportions of
variance in deregressed BLUPs explained by the kinship matrix for each trait (panels) in
each sub-population (colored bars). “NEXTGEN” represents the population-wide
analysis. Values are defined as the ratio of the variance component estimated for the
kinship matrix by TASSEL divided by the total variance in the deregressed-BLUP used
as the response variable in the mixed model implemented by TASSEL.
Population
Trait
Vg
Ve
PVE
NEXTGEN
CMD1S
0.67
0.44
0.60
Genetic Gain
CMD1S
0.37
0.22
0.63
Cycle 1
CMD1S
1.58
0.71
0.69
Cycle 2
CMD1S
0.02
0.04
0.35
NRCRI
CMD1S
0.33
0.26
0.56
NEXTGEN
CMD3S
0.49
0.44
0.53
Genetic Gain
CMD3S
0.36
0.30
0.54
Cycle 1
CMD3S
1.13
0.50
0.69
Cycle 2
CMD3S
0.16
0.35
0.32
NRCRI
CMD3S
0.15
0.21
0.42
NaCRRI
CMD3S
0.54
0.04
0.94
NEXTGEN
CMD6S
0.51
0.37
0.58
Genetic Gain
CMD6S
0.28
0.26
0.52
Cycle 1
CMD6S
0.76
0.43
0.64
NRCRI
CMD6S
0.07
0.16
0.29
NaCRRI
CMD6S
0.64
0.06
0.91
NEXTGEN
CMD9S
0.06
0.10
0.38
NRCRI
CMD9S
0.06
0.10
0.38
NEXTGEN
CMD12S
0.09
0.08
0.54
NRCRI
CMD12S
0.09
0.08
0.54
NEXTGEN
MCMDS
0.36
0.28
0.56
Genetic Gain
MCMDS
0.24
0.20
0.54
Cycle 1
MCMDS
0.89
0.41
0.68
Cycle 2
MCMDS
0.05
0.10
0.31
NRCRI
MCMDS
0.07
0.16
0.30
NaCRRI
MCMDS
0.59
0.03
0.95
NEXTGEN
AUCS
28.53
11.04
0.72
Genetic Gain
AUCS
8.17
5.89
0.58
Cycle 1
AUCS
37.24
11.88
0.76
NRCRI
AUCS
5.74
5.82
0.50
Supplementary Table 10. Pairwise LD between five chromosome 8 markers selected by
multi-locus mixed-model (MLMM).
S8_7762525
S8_6380064
S8_6632472
S8_7325389
S8_4919667
S8_7762525
1.00
0.06
0.05
0.24
0.02
S8_6380064 S8_6632472 S8_7325389 S8_4919667
1.00
0.00
0.03
0.43
1.00
0.01
0.00
1.00
0.00
1.00
Supplementary Figure 1. Distribution of CMD severity scores in the entire collection .
NEXTGEN
25000
20000
MCMDS
15000
10000
5000
0
6000
CMD1S
4000
2000
0
12000
9000
CMD3S
6000
variable
MCMDS
CMD1S
0
CMD3S
CMD6S
7500
CMD9S
CMD6S
5000
2500
0
3000
CMD9S
2000
1000
0
600
CMD12S
400
200
CMD Severity Score
5
4
3
2
0
1
Density
3000
CMD12S
Supplementary Figure 2. Distribution of CMD severity scores in the NaCRRI, Uganda,
training population.
NaCRRI
1500
1000
MCMDS
500
0
1500
CMD3S
Density
1000
500
0
900
CMD6S
600
300
CMD Severity Score
5
4
3
2
1
0
variable
MCMDS
CMD3S
CMD6S
Supplementary Figure 3. Distribution of CMD severity scores in the NRCRI, Nigeria,
training population.
NRCRI
8000
6000
MCMDS
4000
2000
0
1500
CMD1S
1000
500
0
3000
CMD3S
2000
Density
1000
variable
MCMDS
CMD1S
0
CMD3S
CMD6S
4000
CMD9S
CMD6S
3000
2000
1000
0
3000
CMD9S
2000
1000
0
600
CMD12S
400
200
CMD Severity Score
5
4
3
2
1
0
CMD12S
Supplementary Figure 4. Distribution of CMD severity scores in the IITA, Nigeria,
training population (a.k.a. IITA: Genetic Gain).
Genetic Gain
15000
MCMDS
10000
5000
0
2000
CMD1S
1500
1000
500
0
CMD3S
2000
1000
variable
CMD1S
0
4000
CMD3S
CMD6S
CMD9S
3000
CMD6S
2000
1000
0
1250
1000
CMD9S
750
500
250
0
400
CMD12S
300
200
100
CMD Severity Score
5
4
3
2
0
1
Density
MCMDS
CMD12S
Supplementary Figure 5. Distribution of CMD severity scores in the IITA, Nigeria, GS
Cycle 1 progenies.
Cycle 1
3000
2000
MCMDS
1000
0
1000
CMD1S
500
Density
variable
MCMDS
0
CMD1S
3000
CMD3S
CMD6S
2000
CMD3S
1000
0
1000
CMD6S
500
CMD Severity Score
5
4
3
2
1
0
Supplementary Figure 6. Distribution of CMD severity scores in the IITA, Nigeria, GS
Cycle 2 progenies.
Cycle 2
2000
1500
MCMDS
1000
500
0
2500
2000
CMD1S
Density
1500
1000
500
0
2000
1500
CMD3S
1000
500
CMD Severity Score
5
4
3
2
1
0
variable
MCMDS
CMD1S
CMD3S
Supplementary Figure 7. Distribution of de-regressed BLUPs for MCMDS scores in
each sub-population.
MCMDS
NEXTGEN
2000
1000
0
NaCRRI
100
50
0
200
NRCRI
150
100
Density
50
Population
NEXTGEN
NaCRRI
0
NRCRI
GeneticGain
150
GeneticGain
100
50
0
800
Cycle_1
600
400
200
0
2000
Cycle_2
1500
1000
500
Deregressed BLUP for MCMDS
4
2
0
0
Cycle_1
Cycle_2
Supplementary Figure 8. Distribution of de-regressed BLUPs for CMD1S scores in
each sub-population.
CMD1S
2500
2000
NEXTGEN
1500
1000
500
0
125
100
NRCRI
75
50
25
0
125
GeneticGain
Density
100
75
50
25
0
750
Cycle_1
500
250
0
2500
2000
Cycle_2
1500
1000
500
Deregressed BLUP for CMD1S
4
2
0
0
Population
NEXTGEN
NRCRI
GeneticGain
Cycle_1
Cycle_2
Supplementary Figure 9. Distribution of de-regressed BLUPs for CMD3S scores in
each sub-population.
CMD3S
NEXTGEN
2000
1000
0
150
NaCRRI
100
50
0
250
200
NRCRI
150
100
Density
Population
NEXTGEN
50
NaCRRI
0
NRCRI
GeneticGain
GeneticGain
100
50
0
400
Cycle_1
300
200
100
0
2000
Cycle_2
1500
1000
500
Deregressed BLUP for CMD3S
4
2
0
−2
0
Cycle_1
Cycle_2
Supplementary Figure 10. Distribution of de-regressed BLUPs for CMD6S scores in
each sub-population.
CMD6S
1000
750
NEXTGEN
500
250
0
150
NaCRRI
100
50
0
150
NEXTGEN
100
NRCRI
50
NaCRRI
NRCRI
GeneticGain
Cycle_1
0
300
GeneticGain
200
100
0
400
Cycle_1
200
Deregressed BLUP for CMD3S
3
2
1
0
0
−1
Density
Population
Supplementary Figure 11. Distribution of de-regressed BLUPs for CMD9S scores in
each sub-population.
CMD9S
400
300
NEXTGEN
200
100
0
200
150
NRCRI
Density
Population
100
50
0
75
GeneticGain
50
25
Deregressed BLUP for CMD9S
4
3
2
1
0
0
NEXTGEN
NRCRI
GeneticGain
Supplementary Figure 12. Distribution of de-regressed BLUPs for CMD12S scores in
each sub-population.
CMD12S
60
NEXTGEN
40
20
0
40
30
NRCRI
Density
Population
20
10
0
20
GeneticGain
10
Deregressed BLUP for CMD12S
1.0
0.5
0.0
0
NEXTGEN
NRCRI
GeneticGain
Supplementary Figure 13. Distribution of de-regressed BLUPs for AUDPC scores in
each sub-population.
AUCS
600
NEXTGEN
400
200
0
100
75
NRCRI
50
25
Density
Population
NEXTGEN
0
150
NRCRI
GeneticGain
Cycle_1
GeneticGain
100
50
0
400
300
Cycle_1
200
100
20
10
0
Deregressed BLUP for AUCS
Supplementary Figure 14. Linkage-disequilibrium versus physical distance (kb) in the
entire germplasm collection. Raw pair-wise correlations (A) and means of binned
distance classes, by chromosome (B).
A.
B.
NEXTGEN
factor(CHR_A)
1
2
3
4
0.4
5
Bin Mean LD (r2)
6
7
8
9
10
11
0.2
12
13
14
15
16
17
18
Intra−marker Distance Bin (Kilobases)
500 to 1000
250 to 500
100 to 250
50 to 100
25 to 50
10 to 25
4 to 10
2 to 4
1 to 2
0 to 1
0.0
Supplementary Figure 15. Linkage-disequilibrium versus physical distance (kb) in the
NaCRRI, Uganda reference population. Raw pair-wise correlations (A) and means of
binned distance classes, by chromosome (B).
A.
B.
NaCRRI
0.6
factor(CHR_A)
1
2
3
4
5
6
Bin Mean LD (r2)
0.4
7
8
9
10
11
12
0.2
13
14
15
16
17
18
Intra−marker Distance Bin (Kilobases)
500 to 1000
250 to 500
100 to 250
50 to 100
25 to 50
10 to 25
4 to 10
2 to 4
1 to 2
0 to 1
0.0
Supplementary Figure 16. Linkage-disequilibrium versus physical distance (kb) in the
NRCRI, Nigeria reference population. Raw pair-wise correlations (A) and means of
binned distance classes, by chromosome (B).
A.
B.
NRCRI
factor(CHR_A)
0.5
1
2
3
4
0.4
5
Bin Mean LD (r2)
6
7
0.3
8
9
10
11
0.2
12
13
14
15
0.1
16
17
18
Intra−marker Distance Bin (Kilobases)
500 to 1000
250 to 500
100 to 250
50 to 100
25 to 50
10 to 25
4 to 10
2 to 4
1 to 2
0 to 1
0.0
Supplementary Figure 17. Linkage-disequilibrium versus physical distance (kb) in the
IITA, Nigeria reference population (a.k.a. IITA: Genetic Gain). Raw pair-wise
correlations (A) and means of binned distance classes, by chromosome (B).
A.
B.
IITA: Genetic Gain
factor(CHR_A)
1
2
3
4
0.4
5
Bin Mean LD (r2)
6
7
8
9
10
11
12
0.2
13
14
15
16
17
18
Intra−marker Distance Bin (Kilobases)
500 to 1000
250 to 500
100 to 250
50 to 100
25 to 50
10 to 25
4 to 10
2 to 4
1 to 2
0 to 1
0.0
Supplementary Figure 18. Linkage-disequilibrium versus physical distance (kb) in the
IITA, Nigeria GS Cycle 1 progenies. Raw pair-wise correlations (A) and means of binned
distance classes, by chromosome (B).
A.
B.
IITA: Cycle 1
factor(CHR_A)
1
2
3
4
0.4
5
Bin Mean LD (r2)
6
7
8
9
10
11
0.2
12
13
14
15
16
17
18
Intra−marker Distance Bin (Kilobases)
500 to 1000
250 to 500
100 to 250
50 to 100
25 to 50
10 to 25
4 to 10
2 to 4
1 to 2
0 to 1
0.0
Supplementary Figure 19. Linkage-disequilibrium versus physical distance (kb) in the
IITA, Nigeria GS Cycle 2 progenies. Raw pair-wise correlations (A) and means of binned
distance classes, by chromosome (B).
A.
B.
IITA: Cycle 2
0.6
factor(CHR_A)
1
2
3
4
5
0.4
Bin Mean LD (r2)
6
7
8
9
10
11
12
0.2
13
14
15
16
17
18
Intra−marker Distance Bin (Kilobases)
500 to 1000
250 to 500
100 to 250
50 to 100
25 to 50
10 to 25
4 to 10
2 to 4
1 to 2
0 to 1
0.0
Supplementary Figure 20. Quantile-quantile plot comparing observed vs. expected –log10(p-values) for four tested GWAS models
(colors): Naïve GLM, GLM + 5 PCs, GLM + 10 PCs, MLM. Each trait is shown as a panel. Results for the population-wide
(“NEXTGEN”) analysis are shown.
Supplementary Figure 21. Quantile-quantile plot comparing observed vs. expected –log10(p-values) for four tested GWAS models
(colors): Naïve GLM, GLM + 5 PCs, GLM + 10 PCs, MLM. Each trait is shown as a panel. Results for the NaCRRI sub-population
analysis are shown.
Supplementary Figure 22. Quantile-quantile plot comparing observed vs. expected –log10(p-values) for four tested GWAS models
(colors): Naïve GLM, GLM + 5 PCs, GLM + 10 PCs, MLM. Each trait is shown as a panel. Results for the NRCRI sub-population
analysis are shown.
Supplementary Figure 23. Quantile-quantile plot comparing observed vs. expected –log10(p-values) for four tested GWAS models
(colors): Naïve GLM, GLM + 5 PCs, GLM + 10 PCs, MLM. Each trait is shown as a panel. Results for the IITA GS Cycle 1 subpopulation analysis are shown.
Supplementary Figure 24. Quantile-quantile plot comparing observed vs. expected –log10(p-values) for four tested GWAS models
(colors): Naïve GLM, GLM + 5 PCs, GLM + 10 PCs, MLM. Each trait is shown as a panel. Results for the IITA GS Cycle 2 subpopulation analysis are shown.
Supplementary Figure 25. Quantile-quantile plot comparing observed vs. expected –log10(p-values) for four tested GWAS models
(colors): Naïve GLM, GLM + 5 PCs, GLM + 10 PCs, MLM. Each trait is shown as a panel. Results for the IITA Genetic Gain subpopulation analysis are shown.
Supplmentary Figure 26. Examples of main effects at non-chromosome 8 significant markers. Barplots are shown in each panel for a
different marker, representing the mean and standard error for the three marker genotypes.
Supplementary Figure 27. Manhattan plots zoomed in to significance region on
chromosome 8 and broken down by sub-populations (panels). Plot is for association with
mean CMD severity score (MCMDS).
Supplementary Figure 28. (A) Manhattan plot of chromosome 8 showing the locations
of 4 SSR markers previously associated with CMD resistance. (B) Manhattan plot of the
entire genome showing the location of two additional SSR markers (chromosomes 9 and
14) also associated with CMD resistance.
Supplementary Figure 29. Manhattan plot of the significance region on chromosome 8
for MCMDS. The location of haplotype blocks identified by PLINK are shown as bars
below the plot.
Supplementary Figure 30. Manhattan plot showing variance explained (r2) by linear
model tests for interactions between the top marker (S8_7762525, blue vertical line) and
every other maker on chromsome 8.
Supplementary Figure 31. Genome-wide manhattan plot and chromosome 8 zoomed plot (inset) of results from a multi-locus mixed
model. Red dots represent the five markers selected by the model based on BIC as co-factors in the analysis. The red line represents
the Bonferonni significance threshold.
Supplementary Figure 32. Additional examples of two-locus marker genotype effects for the top SNP S8_7762525 and the following
second SNPs: the SNP with the lowest p-value for the interaction-term in post-hoc epistasis tests (S8_4919667, top row) and two other
candidate SNPs, S8_6218789 and S8_9487653 (middle and bottom rows respectively). The left column shows barplots of the mean
and standard error for each multi-locus genotype for the mean CMD severity score. The middle column shows disease progress curves
with the mean and standard error CMD severity across 1, 3 and 6 months after planting. The right column shows the location of the
two markers being compared, S8_7762525 (red line) and the other marker (blue line).
Supplementary Figure 33. Venn diagram of the number of significant SNPs identified per time point: 1,3, 6 and 9 months after
planting.
Supplementary Figure 34. Pie chart depicting the proportion of annotated candidate genes belonging to each of eight gene ontology
categories.
apoptotic process (GO:0006915)
biological regulation (GO:0065007)
cellular component organization or
biogenesis (GO:0071840)
cellular process (GO:0009987)
developmental process
(GO:0032502)
localization (GO:0051179)
metabolic process (GO:0008152)
response to stimulus (GO:0050896)
Supplementary Figure 35. Candidate genes involved in ROS and auxin signaling pathaway within the CMD QTL were scanned for
mutations across different cassava genotypes. Three tolerant genotypes (TMEB3, TMEB7 and I011412) and two susceptible
genotypes (I30572 and TMEB1) were considered. Part of the second exon of Cassava4.1_029175m.g, a peroxidase, is shown. Color
bars show changes of nucleotides when comparing to the reference genome (Blue = C, Brown = G, Green = A, Grey = Reference, Red
= T). Full mutations for the coding regions and their impact on the aminoacid sequence are presented for each gene (Table S3). We
also present xml files from the Integrative Genomics Viewer (IGV) that contain all mutations and indel information for the genomic
regions carrying the genes.