Table 2 Overall effect sizes (95% CIs) of host depletion method based on statistical models

From: Host DNA depletion on frozen human respiratory samples enables successful metagenomic sequencing for microbiome studies

Sample

Treatment

% Host

log10 (Final reads)

Microbial species richness

Viral species richness

Predicted functional richness

Microbial bias (Morisita-Horn)

Viral bias (Morisita-Horn)

% Gram-negative

BAL

lyPMA

−3.1 (−15.7, 9.5)

0.4 (−0.2, 0.9)

1.8 (−10.7, 14.2)

0.4 (−10.7, 14.2)

63.0 (−48.7, 174.7)

0.3 (0.1, 0.6)*

-a

6.4 (−21.2, 34.1)

Benzonase

−1.1 (−13.8, 11.5)

0.8 (0.3, 1.3)*

5.7 (−6.2, 17.6)

5.6 (−6.2, 17.6)

137.2 (25.5, 248.9)*

0.1 (−0.1, 0.3)

-

15.3 (−12.3, 43.0)

HostZERO

−18.3 (−30.9, −5.6)*

1.0 (0.4, 1.5)**

8.9 (−3.0, 20.8)

6.6 (−3.0, 20.8)

177.8 (66.1, 289.5)**

0.3 (0.0, 0.5)

-

8.6 (−19.0, 36.3)

MolYsis

−17.7 (−30.3, −5.1)*

1.0 (0.5, 1.6)**

18.9 (7.0, 30.8)*

15.0 (7.0, 30.8)*

203.2 (91.5, 314.9)***

0.2 (0.0, 0.4)

-

3.4 (−24.3, 31.0)

QIAamp

−6.3 (−18.9, 6.3)

1.0 (0.5, 1.6)**

9.5 (−2.4, 21.4)

10.6 (−2.4, 21.4)

139.4 (27.7, 251.1)*

0.3 (0.0, 0.5)

-

7.3 (−20.4, 35.0)

Nasal

lyPMA

−27.7 (−49.0, −6.3)*

−0.5 (−1.0, −0.1)*

−4.8 (−9.2, −0.4)

−6.4 (−9.2, −0.4)*

7.8 (−26.7, 42.3)

0.2 (0.1, 0.3)*

0.5 (0.1, 0.3)***

19.4 (14.1, 24.6)***

Benzonase

−20.0 (−41.4, 1.5)

0.1 (−0.3, 0.6)

−0.4 (−4.8, 4.1)

−3.9 (−4.8, 4.1)

23.4 (−11.1, 57.9)

0.2 (0.0, 0.3)*

0.1 (0.0, 0.3)

1.9 (−3.4, 7.5)

HostZERO

−73.6 (−94.9, −52.1)***

0.9 (0.4, 1.3)**

10.0 (5.6, 14.5)***

4.1 (5.6, 14.5)

69.4 (34.9, 103.9)***

0.0 (−0.1, 0.2)

0.1 (−0.1, 0.2)

0.0 (−5.3, 5.6)

MolYsis

−50.6 (−72.0, −29.3)***

0.2 (−0.2, 0.7)

6.2 (1.8, 10.6)*

−0.2 (1.8, 10.6)

56.2 (21.7, 90.7)**

0.2 (0.1, 0.3)**

0.2 (0.1, 0.3)

2.3 (−3.0, 7.6)

QIAamp

−75.4 (−96.9, −54.0)***

1.1 (0.6, 1.5)***

7.8 (3.3, 12.2)**

4.1 (3.3, 12.2)

69.4 (34.9, 103.9)***

0.1 (0.0, 0.3)

0.4 (0.0, 0.3)**

0.1 (−5.4, 5.5)

Sputum

lyPMA

−3.8 (−15.4, 7.8)

0.5 (0.3, 0.8)**

37.6 (19.5, 55.7)**

4.2 (19.5, 55.7)

86.0 (19.6, 152.4)*

0.3 (0.1, 0.6)**

0.2 (0.1, 0.6)

−40.9 (−52.6, −29.1)***

Benzonase

−6.3 (−17.9, 5.4)

0.8 (0.6, 1.1)***

66.6 (48.5, 84.7)***

19.2 (48.5, 84.7)

91.6 (25.2, 158.0)**

0.5 (0.3, 0.7)***

0.3 (0.3, 0.7)

−52.5 (−64.2, −40.7)***

HostZERO

−45.5 (−57.1, −33.8)***

1.7 (1.4, 1.9)***

103.0 (84.9, 121.1)***

91.8 (84.9, 121.1)***

145.8 (79.4, 212.2)***

0.6 (0.4, 0.8)***

0.7 (0.4, 0.8)**

−59.9 (−71.6, −48.1)***

MolYsis

−69.6 (−81.3, −58.0)***

2.0 (1.7, 2.2)***

112.8 (94.7, 130.9)***

118.4 (94.7, 130.9)***

150.4 (84.0, 216.8)***

0.6 (0.4, 0.8)***

0.8 (0.4, 0.8)***

−59.9 (−71.6, −48.1)***

QIAamp

−18.7 (−30.3, −7.1)**

1.4 (1.2, 1.7)***

85.2 (67.1, 103.3)***

46.6 (67.1, 103.3)*

123.6 (57.2, 190.0)***

0.6 (0.4, 0.8)***

0.7 (0.4, 0.8)**

−60.6 (−72.3, −48.8)***

  1. Changes were assessed using linear mixed-effect models (features ~ host depletion treatment + (1|subject)) stratified by sample type; untreated samples modeled as the comparator group. Sample sizes were n = 30, 35 and 30 for BAL, nasal, and sputum samples, respectively. For microbial bias, samples with no reads were omitted from the analysis, and sample sizes were n = 28, 35, and 30 for BAL, nasal, and sputum samples, respectively. For viral bias, samples with no reads were omitted from the analysis, and sample sizes were n = 35 and 29 for nasal and sputum samples, respectively. *p-value < 0.05, **p-value < 0.01 and ***p-value < 0.001. Values depicted effect size (95% confidence intervals) derived from statistical models. With BAL, MolYsis was the optimal method. lyPMA, Benzonase, and QIAamp showed low depletion efficiency, HostZERO did not increase richness. With nasal swabs, QIAamp was the optimal method. lyPMA and Benzonase showed no microbial richness increase, HostZERO showed high library prep failure rate, MolYsis showed high bias. With sputum HostZERO was the optimal method. lyPMA and Benzonase showed low depletion efficiency, MolYsis and QIAamp showed higher bias.
  2. aNote the majority of BAL samples had no viral species identified thus unable to perform statistical modeling.
Back to article page