Accurate diagnosis of atopic dermatitis by combining transcriptome and microbiota data with supervised machine learning

Abstract

Atopic dermatitis (AD) is a common skin disease in childhood whose diagnosis requires expertise in dermatology. Recent studies have indicated that host genes-microbial interactions in the gut contribute to human diseases including AD. We sought to develop an accurate and automated pipeline for AD diagnosis based on transcriptome and microbiota data. Using these data of 161 subjects including AD patients and healthy controls, we trained a machine learning classifier to predict the risk of AD. We found that the classifier could accurately differentiate subjects with AD and healthy individuals based on the omics data with an average F1-score of 0.84. With this classifier, we also identified a set of 35 genes and 50 microbiota features that are predictive for AD. Among the selected features, we discovered at least three genes and three microorganisms directly or indirectly associated with AD. Although further replications in other cohorts are needed, our findings suggest that these genes and microbiota features may provide novel biological insights and may be developed into useful biomarkers of AD prediction.

Figure 1

The overview of atopic dermatitis classification pipelines in two settings. (a) Transcriptome dataset only, and (b) transcriptome and microbiota data.

Figure 2

The ROC curve of the test set with transcriptome data only. (a) All features (44,608) + SVM (rbf). (b) Chi-squared test (35) + SVM (rbf). (c) All features (44,608) + SVM (rbf), with noise (I = 0.001) and probability threshold = 0.3. (d) Chi-squared test (35) + SVM (rbf), with noise (I = 0.001) and probability threshold = 0.3.

Figure 3

The ROC curve of the test set with microbiota data. (a) All features (366) + SVM (rbf). (b) Chi-squared test (25) + SVM (rbf). (c) Chi-squared test (85) + SVM (rbf). (d) Chi-squared test (85) + SVM (rbf), with noise (I = 0.001) and probability threshold = 0.3. For panel (a,b), we only use microbiota data, while for (c,d) we also include transcriptome data.

Figure 4

The average feature importance of the top 35 selected probes/genes. See more detailed annotation information in Supplementary Table S5.

Figure 5

The average feature importance of the top 50 selected microorganisms from the microbiota dataset.