1
Quality score based identification and correction of pyrosequencing errors
2
Shyamala Iyer, Heather Bouzek, Wenjie Deng, Brendan Larsen, Eleanor Casey,
3
and James I. Mullins
4
5
6
7
Supplementary Data
Supplementary Table 1. Effect of varying coverage threshold on sensitivity
8
and specificity of SNP variant calling. AmpliconNoise + CorQ error correction
9
was used on pyrosequences mapping to the env region (~2500nt) from the ten
10
HIV-1 genome control dataset. Different fold coverage values were used as input
11
parameters in CorQ. Sensitivity and specificity of SNP variant detection within
12
this region is calculated for each fold coverage value.
13
14
Supplementary Table 2. Average number of reads and average read length
15
for simulated pyrosequences. Two sets of simulated pyrosequences generated
16
using Flowsim [1] are shown here. The first set (Set 1a, b and c) is comprised of
17
simulated reads generated using a single 1500 nt HIV-1 sequence as the starting
18
template. The second set (Set 2a, b and c) is comprised of simulated reads
19
generated using a 1500nt region located within 28 HIV-1 sequences as starting
20
templates. Simulations were done without additional SNP errors (1a, 2a) and with
21
two different SNP error rates, 0.005 and 0.01 (1b,c and 2b,c).
22
23
Supplementary Table 3. Comparison of insertion, deletion and substitution
24
error rates in homopolymeric regions after error correction on simulated
25
pyrosequences. Simulated reads were generated using Flowsim [1] using a
26
single 1500nt HIV-1 sequence as the starting template (Simulated datasets 1a-
27
c). Average insertion, deletion and substitution error rates within homopolymeric
28
regions are shown after correction with no additional SNP errors, and SNP error
29
rates of 0.005 and 0.01.
30
31
Supplementary Table 4. Comparison of insertion, deletion and substitution
32
error rates in non-homopolymeric regions after error correction on
33
simulated pyrosequences. The simulated reads were generated in Flowsim [1]
34
using a single 1500nt HIV-1 sequence as the starting template (Simulated
35
datasets 1a-c). Average insertion, deletion and substitution error rates within
36
non-homopolymeric regions are shown after correction with no additional SNP
37
errors, and SNP error rates of 0.005 and 0.01.
38
39
Supplementary Table 5. Sensitivity and specificity comparison of error
40
correction and SNP calling algorithms on simulated pyrosequences.
41
Simulated datasets 2a-c was used to compare the sensitivity and specificity of
42
error correction algorithms. Sensitivity measures the proportion of true SNPs
43
present within the HIV-1 templates, and correctly identified as such by the
44
various SNP calling programs. Specificity measures the proportion of true
45
negatives (positions in the gene regions that are invariant) that are correctly
46
identified as such by the compared programs. Note that QuRe failed when used
47
on simulated pyrosequences generated with a SNP error rate of 0.005. * Values
48
from QuRe are shown when the poor coverage regions were excluded from
49
sensitivity analysis and when these regions are included as false negatives
50
during analysis (the latter values are shown in parenthesis).
51
52
Supplementary Figure 1. Average reduction in base quality for indels found
53
in homopolymer and non-hompolymer regions. Reduction in base quality was
54
measured as the average difference in quality between flagged positions with
55
indels and the adjacent columns (See Materials and Methods, Equation 1). Base
56
qualities from uncorrected sequences (raw 454), and sequences corrected with
57
AmpliconNoise and Pyrobayes are shown for indels found in non-homopolymer
58
regions (length of 1) and varying homopolymer lengths. Reduction in base quality
59
is shown for indels within gag (A), env (B), nef (C) and the three genes combined
60
(D).
61
62
Supplementary File 1
63
Sanger sequences used as templates to generate simulated pyrosequencing
64
sets (2a-c) [2].
65
References
66
67
68
69
70
71
72
73
1. Balzer S, Malde K, Lanzen A, Sharma A, Jonassen I (2010) Characteristics of
454 pyrosequencing data--enabling realistic simulation with flowsim.
Bioinformatics 26: i420-425.
2. Herbeck JT, Rolland M, Liu Y, McLaughlin S, McNevin J, et al. (2011)
Demographic processes affect HIV-1 evolution in primary infection before
the onset of selective processes. Journal of virology 85: 7523-7534.