Background: Osteopenia and osteoporosis are both common bone diseases in postmenopausal women, but their clinical manifestations tend to be unnoticeable until bone loss-induced fractures occur which could endanger personal safety and lead to other injuries. This phenomenon makes it difficult to identify the population with osteoporosis, leading to failure to receive treatment timely. As a significant public health concern, the early diagnosis of osteopenia and osteoporosis has become an urgent clinical topic.

Methods: We conducted metabolomic analysis on plasma samples and 16S sequencing on fecal samples from humans. While 142 women participated in metabolomic analysis, only 118 of them joined the transcriptomics cohort. By statistical methods, we managed to identify relevant biomarkers of osteopenia and osteoporosis consisting of metabolites and genera and establish early diagnosis models, including clinical features, classical indexes and selected biomarkers. The efficiency of these models was also validated.

Results: Metabolites included homo-L-arginine, 3-methylglutarylcarnitine, linoleyl carnitine, and estrone glucuronide, and gut microbes included Coprococcus, Dialister, Erysipelatoclostridium, Marvinbryantia and Stenotrophomonas were identified as biomarkers of osteoporosis while the metabolite dihomo-gamma-linolenoylethanolamide and genera including Lachnospiraceae_NC2004_group, Lachnospiraceae_NK4A136_group and Roseburia were identified as as biomarkers of osteopenia. The area under the curve (AUC) of early diagnosis models for distinguishing osteoporosis from normal, osteopenia from normal and osteoporosis from osteopenia were 0.95, 0.77, and 0.92 respectively.

Conclusion: We identified biomarkers of both osteopenia and osteoporosis consisting of metabolites and gut microbes and investigated the mechanisms underlying their influence on the reduction of BMD. High-reliability prediction models were built through artificial neural networks and random forest and these models can help early diagnosis of osteopenia and osteoporosis to realize early intervention and early treatment.

The fecal samples were collected in sterile plastic cups from participants and stored at -80 °C within 1 hour. The microbial DNA was extracted by using TIANamp Stool DNA Kit (TIANGEN, Beijing, China) and the DNA solution was stored at -20°C. The concentration of the extracted DNA was determined by Nanodrop (Spectrophotometer ND1000) for further amplification. The forward Illumine adapter sequence was CCTACGGGNGGCWGCAG and the reverse was GACTACHVGGGTATCTAATCC.

The sequencing results were analyzed according to Qiime2 procedure. Illumina NovaSeq 6000 was used for acquiring the raw sequencing data and DADA2 was applied for filtration and denoising of the raw data which is equivalent to clustering DNA sequences with 10% similarity[26].  Only low-quality sequences were removed and corrected, and algorithm recognition was used to remove chimeras. The redundancy of the denoised sequences was directly removed redundancy to obtain features, including OTUs, ASVs, and other information.

The clear data were acquired from raw data by filtration treatment. By deleting reads with low-quality bases of mass value data ≤14 exceeding, N bases ³ 10 or overlap between adaptor and reads exceeding 10, the filtration result is presented in Figure S2A. The sequencing error rate was used to detect abnormal base positions with high error rates within the sequencing range (Figure S2B). The quality of the sequencing data was mainly distributed above Q30 (≥ 80%) to ensure the normal progression of the subsequent analysis. According to the characteristics of sequencing technology, the base quality at the end of sequencing reads is generally lower than that at the front end.