Table 2 The topological features of the microbial connection network according to the exposure levels of food emulsifiers

From: Distinct diet-microbiome associations in autism spectrum disorder

Polysorbate 80

non-ASD

ASD

Parameters

Low (n = 143)

Moderate (n = 115)

High (n = 98)

Low (n = 130)

Moderate (n = 157)

High (n = 175)

Number of edges

85bc

111bc

119c

76b

57b

45a

Number of positive edges

76bc

105bc

113c

68b

49b

38a

Number of negative edges

9c

6a

6a

8c

8c

7b

Number of vertices

59b

70c

75c

47a

53a

39a

Average degree

2.881ab

3.171ab

3.173b

3.234b

2.151ab

2.307a

Average path length

3.658a

3.679a

5.794a

4.303a

2.425a

2.507a

Clustering coefficient

0.426ab

0.458ab

0.428ab

0.558b

0.333a

0.361a

Centralization degree

7.241a

7.942a

8.946a

7.934a

8a

7.894a

Modularity (> 0.4)

0.626a

0.614a

0.584a

0.45a

0.736a

0.626a

Number of modules

11ab

13ab

14ab

10ab

14b

9a

Carrageenan

non-ASD

ASD

Parameters

Low (n = 142)

Moderate (n = 132)

High (n = 82)

Low (n = 131)

Moderate (n = 140)

High (n = 191)

Number of edges

67bc

119bc

79c

69b

70bc

32a

Number of positive edges

64bc

107bc

74c

64b

64bc

24a

Number of negative edges

3a

12c

5b

5b

6b

8b

Number of vertices

51b

73bd

57c

43b

61b

31a

Average degree

2.627bc

3.26c

2.772c

3.209b

2.295b

2.064a

Average path length

4.965a

5.212a

2.193a

2.173a

2.761a

2.799a

Clustering coefficient

0.574a

0.47a

0.672a

0.594a

0.275a

0.328a

Centralization degree

6.441a

9.875a

7.357a

9a

5.8a

5.1a

Modularity (> 0.4)

0.684a

0.637a

0.672b

0.506a

0.738ab

0.658a

Number of modules

10ab

13ab

15c

8b

12a

7a

  1. In the absence of standard limits for annual intake of food emulsifiers, we classified the general population into tertiles, categorized as low, moderate, and high exposure groups. The topological features of the microbial connection network were analyzed separately in children with and without ASD using the igraphR package. To estimate the variability of network metrics, we employed a nonparametric bootstrap procedure with 1000 iterations via repeated resampling with replacement. Two-tailed, bias-corrected p-values were computed for the statistical differences of each topological feature. The compact letter (a, b, c) display denoted significant differences between groups, with those not sharing any common letter determined as significantly different after multiple comparisons (p < 0.05).