Myasthenia gravis (MG) is a neuromuscular junction disease with a complex pathophysiology and clinical variation for which no clear biomarker has been discovered. We hypothesized that because changes in gut microbiome composition often occur in autoimmune diseases, the gut microbiome structures of patients with MG would differ from those without, and supervised machine learning (ML) analysis strategy could be trained using data from gut microbiota for diagnostic screening of MG. Genomic DNA from the stool samples of MG and those without were collected and used to establishe a sequencing library by constructing amplicon sequence variants (ASVs) and completing taxonomic classification of each representative DNA sequence. Four ML methods with nested leave-one-out cross-validation were trained using ASV taxon–based data and full ASV–based data to identify key ASVs in each data set. Overlapping key features extracted when XGBoost was trained using the full ASV–based and ASV taxon–based data were identified, and 31 high-importance ASVs (HIASVs) were obtained. The most significant difference observed was in the abundance of bacteria in the Lachnospiraceae and Ruminococcaceae families. The 31 HIASVs were used to train the XGBoost algorithm to differentiate individuals with and without MG. The model had high diagnostic classification power and could accurately predict and identify patients with MG. In addition, the abundance of Lachnospiraceae was associated with limb weakness severity. 

In this prospective study, 19 individuals with MG and 10 individuals without were consecutively recruited from Fu-Jen Catholic University Hospital. Individuals were enrolled in the MG group if they 1) were given a diagnosis of MG on the basis of having the combination of symptoms and signs that are characteristic of muscle weakness with diurnal changes and either 2a) had a positive test result for specific autoantibodies or 2b) had a positive electrophysiological diagnosis obtained using single-fiber electromyography and repetitive nerve stimulation (Rousseff, 2021). None of the participants had received any abdominal chirurgic intervention; consumed antibiotics, probiotics, or antacids during the previous 6 months; or reported gastrointestinal symptoms during the previous year. This study was approved by the Regional Ethics Committee of Fu-Jen Catholic University Hospital and written informed consent was obtained from each participant (No. FJUH109043). All experiments were completed in accordance with the Declaration of Helsinki’s Ethical Principles for Medical Research Involving Human Subjects and under a set of approved guidelines and regulations. The severity of MG was determined using quantitative MG (QMG), MG activities of daily living (MG-ADL), MG composition (MGC), and MG quality of life (MG-QoL) scores (Jaretzki et al., 2000). Using the categories of the QMG and MGC scales, we categorized the scores on these scales into ocular, bulbar, and limb groups. Fresh stool samples were collected from each participant and immediately frozen at −80°C until analysis. Figure 1 summarizes the overall study workflow.

Each stage in the process, including the sample testing and polymerase chain reaction (PCR) and library creation and sequencing, can affect the quality of the data, and the accuracy of analytical findings is directly influenced by the quality of data. Therefore, quality control measures were implemented at each stage of the process to ensure data accuracy.

DNA Extraction and 16S Metagenomics Sequencing

Genomic DNA was extracted from the samples using the EasyPrep Stool Genomic DNA Kit (Biotools, New Taipei City, Taiwan). The DNA concentration was determined and adjusted to 5 ng/μL for subsequent processing. In accordance with the 16S Metagenomic Sequencing Library Preparation protocol (Illumina), the specific primer set 341F: 5’-CCTACGGGNGGCWGCAG-3’, 806R: 5’-GACTACHVGGGTATCTAATCC-3’ was employed to amplify the variable regions V3 and V4 of the 16S rRNA gene. A PCR was conducted using KAPA HiFi HotStart ReadyMix (Roche) and 12.5 ng of genomic DNA (gDNA) under the following conditions: 95°C for 3 min, 25 cycles of 95°C for 30 s, 55°C for 30 s, 72°C for 30 s, and a final extension of 72°C for 5 min. The reaction was subsequently maintained at 4°C. The products of the PCR were evaluated using 1.5% agarose gel, and samples with a bright main strip at approximately 500 bp were selected for further library preparation. The selected samples were purified using AMPure XP beads.

A sequencing library was prepared using the 16S Metagenomic Sequencing Library Preparation procedure (Illumina). To summarize, the 16S rRNA V3–V4 region PCR amplicon was subjected to a secondary PCR, which was conducted using the Nextera XT Index Kit with dual indices and Illumina sequencing adapters from Illumina. The indexed PCR product was evaluated for quality by using the Qubit 4.0 Fluorometer (Thermo Scientific) and a Qsep100TM system. The indexed PCR products were mixed in equal amounts to create a sequencing library. The library was sequenced on an Illumina MiSeq platform, which generated 300‐bp paired reads.