Illuminating the oral microbiome and its host interactions: recent advancements in omics and bioinformatics technologies in the context of oral microbiome research

Abstract

The oral microbiota has an enormous impact on human health, with oral dysbiosis now linked to many oral and systemic diseases. Recent advancements in sequencing, mass spectrometry, bioinformatics, computational biology, and machine learning are revolutionizing oral microbiome research, enabling analysis at an unprecedented scale and level of resolution using omics approaches. This review contains a comprehensive perspective of the current state-of-the-art tools available to perform genomics, metagenomics, phylogenomics, pangenomics, transcriptomics, proteomics, metabolomics, lipidomics, and multi-omics analysis on (all) microbiomes, and then provides examples of how the techniques have been applied to research of the oral microbiome, specifically. Key findings of these studies and remaining challenges for the field are highlighted. Although the methods discussed here are placed in the context of their contributions to oral microbiome research specifically, they are pertinent to the study of any microbiome, and the intended audience of this includes researchers would simply like to get an introduction to microbial omics and/or an update on the latest omics methods. Continued research of the oral microbiota using omics approaches is crucial and will lead to dramatic improvements in human health, longevity, and quality of life.

This review provides a comprehensive perspective of the current state-of-the-art tools available to perform genomics, metagenomics, phylogenomics, pangenomics, transcriptomics, proteomics, metabolomics, lipidomics, and multi-omics analysis on (all) microbiomes, and then provides examples of how the techniques have been applied to research of the oral microbiome, specifically.

Introduction

The oral microbiota is a unique and diverse community of bacteria, viruses, fungi, and archaea that plays a major role in human health (Baker et al. 2017). Distinct microenvironments within the oral cavity, such as the hard surface of the tooth, keratinized hard palate, or soft surface of the tongue, result in the establishment of unique and highly structured communities at each site (Human Microbiome Project 2012, Lamont et al. 2018). The health-associated oral microbiota exhibits colonization resistance and plays an active role in preventing dysbiosis and associated disease (He et al. 2014, Radaic and Kapila 2021). Meanwhile, dysbiosis of the oral microbiome, even on a highly localized scale, is responsible for dental caries and periodontal disease, both extremely prevalent and costly (Bowen et al. 2018). Furthermore, the majority of oral cancers are driven by oral infection with viruses such as human papilloma virus (HPV) and Epstein-Barr virus (EBV, formerly known as human gammaherpesvirus 4/HHV-4) (Tsao et al. 2017, Economopoulou et al. 2020). In addition to oral diseases, there are increasing lines of evidence linking the oral microbiota to a myriad of extra-oral and systemic diseases, such as obesity, diabetes, cardiovascular disease, inflammatory bowel disease, nonalcoholic fatty liver disease, rheumatoid arthritis, colorectal cancers, and Alzheimer’s disease (Hajishengallis and Chavakis 2021). The oral microbiome has also served as an important model system for researching microbiomes broadly, as diverse taxa across all kingdoms of life co-exist and interact at a site that is easily accessible to observe the processes of biofilm and community assembly and succession (Baker et al. 2017).

Despite significant progress in our understanding of the human oral microbiota, continued research is essential and will lead to improvements in human health and overall quality of life.

Prior to the development of culture-independent analysis methods such as untargeted (i.e. “shotgun”) sequencing and mass spectrometry (MS), the study of the oral microbiome and its role in human health was limited to taxa that could be isolated and cultivated in the laboratory. Using these classic microbiological techniques, key members of the community, including both pathogens (e.g. Streptococcus mutans and Porphyromonas gingivalis) and commensals (e.g. S. gordonii and S. sanguinis) were discovered, became well-studied, and mechanisms of caries and periodontal disease pathogenesis were elucidated. However, the overall picture of the oral microbiota (and indeed all microbiomes) and its role in human health was still relatively incomplete and had a very narrow focus.

Over the past 20 years, culture-independent analysis methods have enabled the formation and subsequent explosive growth of microbiome research, including that of the human oral microbiome. The development of these methods was due to major advancements in sequencing technology, MS, bioinformatics, computational biology, and computer science/machine learning. In concert with the development of microbiome research has been the development of the omics fields of study. Especially pertinent to microbiome research are genomics, metagenomics, phylogenomics, pangenomics, and transcriptomics, which are based on nucleic acid sequencing, as well as metabolomics, proteomics, and lipidomics, which are based on MS. Traditional omics analyzes populations of cells within samples in aggregate, getting an average for the population, which may not reflect the true profiles of a given analyte across individual cells in the population. Single-cell analysis techniques are rapidly addressing this issue, already becoming a mainstay in eukaryotic transcriptomics. Single-cell analysis is much more challenging in bacteria, as cells and therefore the amount of input material are orders of magnitude smaller. However, the first single-cell analyses of bacteria have been described in the past several years. Meanwhile, multi-omics research, examining datasets from two or more omics fields, presents great potential for new discovery but also additional challenges. The continued evolution of these fields of research has enabled the study of the oral microbiome at an unprecedented scale and level of resolution. This review will provide an overview of these omics disciplines and explain some of the most used and state-of-the-art technologies and techniques. The review will then discuss how these approaches have been applied to the study of the oral microbiome, highlighting some of the major recent discoveries that have been facilitated. Since several recent reviews have excellently summarized the use of omics techniques in both dental caries (Bostanci et al. 2021, Moussa et al. 2022) and periodontal disease (Nguyen et al. 2020, Bostanci et al. 2021, Kumar et al. 2021) research, this review will focus more on the omics techniques and tools themselves, including historical context and the current state of the technology. While it is not possible to include all of the technologies, tools, and research worthy of inclusion, this review provides the reader with reference to further comprehensive reviews on more specific topics where possible. This review will also go into more depth on sequencing-based omics rather than MS-based omics, mainly because the former approaches have been more extensively employed by the field of microbiome research.

Sequencing-based omics

Historical background: next-generation sequencing (NGS) revolutionizes the life sciences and enables early microbiome research in the 2000s and early 2010s

NGS methods, including sequencing-by-synthesis (Illumina), pyrosequencing (454 Life Sciences), and sequencing by oligonucleotide ligation and detection (SOLiD; Applied Biosystems), revolutionized the life sciences in the 2000s and early 2010s by enabling accurate, high-throughput, untargeted sequencing (Bennett 2004, Margulies et al. 2005, Bentley et al. 2008, McKernan et al. 2009). For the first time, microbiological samples could be analyzed for all microbial DNA or RNA content, regardless of the cultivability of the taxa present (Venter et al. 2004, Ley et al. 2005, Gill et al. 2006). This led to the establishment of microbiome research as a scientific field and the subsequent explosion of microbiome studies, including large, concerted efforts such as the Human Microbiome Project (Human Microbiome Project 2012). The vast majority of this early microbiome research was conducted using amplicon sequencing-based analysis methods, largely of the 16S rRNA gene (termed “16S sequencing” or “16S analysis”). This was because 16S analysis allows many more samples to be analyzed with a sufficient depth to acquire microbiome data on a sequencing run compared to metagenomics sequencing. As a result, 16S sequencing is higher throughput and significantly cheaper on a per-sample basis. It is important to note that advancements during this period were not limited to sequencing instrumentation and that there were also major developments in MS, computer science, and computational biology that were foundational to many of the modern technologies discussed in this review. Among these were the algorithms and suites of analysis tools that were the first versions and/or predecessors of some of the tools still most widely used in microbiome studies today, including the precursors to the DADA2 (Callahan et al. 2016), QIIME2 (Bolyen et al. 2019), Kraken (Lu et al. 2022), bioBakery (Beghini et al. 2021), SEQUEST (Brodbelt and Russell 2015), and SPAdes (Prjibelski et al. 2020) algorithms and suites of software.

A notable advance in the study of the oral microbiome, specifically, during this period was the development of the Human Oral Microbiome Database (HOMD), first published in 2010 (Chen et al. 2010). This not only provided a free, public, large-scale database of 16S rRNA sequences specific to microbes from the human oral cavity but also began to illustrate how limited previous understanding of the oral microbiota had been, highlighting that 53% of the 619 species-level taxa identified in the project had not been properly named and 35% had never been isolated or cultivated (Chen et al. 2010). The Human Microbiome Project also significantly advanced our understanding of the inhabitants of the oral microbiome at specific niches (Human Microbiome Project 2012). Figure 1A is a timeline illustrating many of the major milestones in omics, microbiome, and oral microbiome research over the last several decades.

Figure 1.

(A) Timeline of milestones in omics technologies and oral microbiome research. This timeline highlights milestones in microbiome/omics research generally (yellow), major technological advances (green), and milestones in oral microbiome research specifically (blue) over the past 33 years. (B) Omics approaches and tools. For each of the seven omics approaches discussed here, a list of the most significant and/or commonly used bioinformatics tools is provided. Note that this list is not exhaustive, and readers are referred in the main text to additional references on the specific software and benchmarking. **denotes a particularly useful or “gold standard” tool. ONT, Oxford Nanopore Technologies.

Current developments sequencing technologies

In the present day, new advancements in sequencing technology are in the process of revolutionizing microbiome research once more. Throughout the 2010s, Illumina emerged as the dominant player in sequencing, holding about 80% of the market share as of 2020, with improvements to their sequencing-by-synthesis technology increasing throughput dramatically while greatly reducing the cost of sequencing. This decrease in sequencing cost has even eclipsed Moore’s Law (which posited that the number of transistors on an integrated circuit doubles about every two years, therefore dropping the cost of computer power to the consumer in a log-linear manner), with the cost of sequencing one million base pairs falling from $10 million in 2001 to $0.10 by 2016 (Wetterstrand 2023, Muir et al. 2016). Interestingly, this phenomenon has led some scientists to hypothesize that computing power and storage will ultimately become the limiting cost factors in sequencing-based research rather than the sequencing itself (Muir et al. 2016). This dramatic reduction in sequencing cost has enabled many more oral health and microbiome researchers to perform larger-scale 16S sequencing projects, metagenomics, whole genome sequencing, and RNA-seq.

At the same time, emerging third-generation sequencing technologies, especially long-read technologies such as nanopore sequencing (Oxford Nanopore [ONT]) (Jain et al. 2015), single-molecule real-time sequencing (SMRT; Pacific Biosciences [PacBio]) (Roberts et al. 2013), and LoopSeq (Element Biosciences) (Callahan et al. 2021), are in the process of transforming the landscape of sequencing yet again and are challenging Illumina’s preeminence. Although Illumina sequencing is highly accurate, the reads produced typically only 150 or 300 bp in length. With read lengths this short, repeat and nonspecific regions significantly hamper efforts to assemble complete genomes, with Illumina-based genome (or metagenome) assemblies typically being split into fragments, which are called contigs (Athanasopoulou et al. 2021). Using ONT sequencing, the length of reads produced is theoretically only limited by the length of the input material, and single reads of over 1 mbp are now routinely reported (Jain et al. 2015). These long reads span the entirety of repeat regions, enabling the assembly of circular chromosomes and complete genomes with much greater ease. RNA can also be sequenced using ONT, where sequencing of the full-length transcripts easily provides transcriptome-wide information on co-transcribed genes and the identification of novel RNA isoforms (Garalde et al. 2018). In addition, ONT sequencing can sequence native molecules, reducing bias by sidestepping the PCR and/or cDNA synthesis steps that are required in many sequencing library preparation protocols (Garalde et al. 2018). Crucially, sequencing native molecules also enables the detection of base modifications and noncanonical bases (e.g. methylated bases, inosine, pseudouridine, etc.), allowing these phenomena to be studied on a genome, metagenome, transciptome, or metatranscriptome scale for the first time (Garalde et al. 2018). These epigenetic modifications have been particularly understudied in the context of microbiology. The most substantial drawback to ONT sequencing is a relatively low accuracy. Errors in ONT sequencing are not random but usually occur during homopolymeric tracts, where the basecalling software has difficulty identifying how many consecutive iterations of a given base or bases have passed through the nanopore, as the rate of processivity through the channel is saltatory, not constant (Amarasinghe et al. 2020). This leads to insertions or deletions, which are nontrivial as they are likely to cause apparent frameshifts and therefore impact downstream gene calling and annotation (Watson and Warr 2019). As a result, ONT sequencing data is frequently combined with Illumina sequencing data of the same sample, where the long reads enable accurate large-scale assembly of contigs and scaffolds, and the short reads are used to polish out the errors inherent to the ONT reads (Koren et al. 2012). Crucially, the accuracy of ONT sequencing has rapidly improved in recent years, falling from 30%–40% in 2015 to <0.1% in raw reads (or <0.001% in a consensus assembly with ≥20X coverage) using current instrumentation and software (Sereika et al. 2022). As a result of these recent, massive improvements in accuracy, several recent studies have shown that the field is reaching an inflection point where accurate genomics and metagenomics can be performed using ONT sequencing alone (Faulk 2022, Liu et al. 2022b, Sereika et al. 2022).

SMRT sequencing technology from PacBio represents a “middle ground” between Illumina and ONT sequencing technologies, combining relatively long reads, averaging 10–25 kb, and an error rate of <0.1% for raw reads and <0.003% for 25–30X consensus assemblies (Wenger et al. 2019). While SMRT sequencing was able to produce accurate genomes and metagenomes independently of short-read polishing much earlier than nanopore sequencing, the significantly higher cost per base of PacBio sequencing and the much higher cost of the PacBio sequencing machines themselves have remained a barrier for many researchers (Sereika et al. 2022). Like ONT sequencing, PacBio sequencing can also sequence full-length RNAs (Leung et al. 2021) and can detect methylation, enabling genome-wide epigenetic studies (Beaulaurier et al. 2018). In addition to Oxford Nanopore and PacBio, newcomers in the sequencing space, such as Element Biosciences (developing both innovations to short-read sequencing and long read LoopSeq) and Stratos Genomics (now owned by Roche, developing sequencing-by-expansion technologies), may indeed further disrupt the industry. In addition, synthetic long-read and linked-read approaches, such as TELL-seq, use labeling of short-reads adjacent on the genome to obtain contiguity of short-read-based assemblies similar to those obtained through long-read-based approaches (Wang et al. 2019). However, limitations, including incompatibility with metagenomic assemblies, continue to limit widespread use of these approaches (Wang et al. 2019).

Genomics

Figure 1B provides a list of bioinformatics tools, and their references, that are discussed in the following sections. Obtaining genomes that are both complete (i.e. contiguous chromosomes and plasmids) and accurate is of prime importance to microbiology research. High-quality, complete genomes (assuming they are publicly available to researchers in a database) enable: (1) accurate detection and quantification of a particular taxon, or its RNA transcripts, in an isolate or microbiome sample (Venter et al. 2004), (2) prediction of the metabolic pathways and therefore possible ecological and pathogenic roles of the taxon—particularly important for taxa that have not yet been isolated or cultivated (Naito et al. 2016), and (3) guiding wet-lab research, such as mutagenesis. It is important to recognize that many genomes in public repositories were assembled using short-read sequencing only, meaning that they are probably at an incomplete or draft stage and fragmented into contigs of various numbers and sizes. These genomes are likely to be missing sequences and may contain contaminant contigs. Therefore, it is crucial that researchers are cognizant of the limitations inherent with these assemblies if they are used as a reference.

To obtain a genome, sequencing reads that have passed quality control must be assembled. Note that the assembly of multispecies (i.e. microbiome) samples is discussed in the following section on metagenomics. A range of assembly tools and algorithms are available to assemble microbial genomes. For Illumina short reads, these include ABySS (Simpson et al. 2009), Velvet (Zerbino and Birney 2008), MEGAHIT (Li et al. 2015), and SPAdes (Prjibelski et al. 2020). SPAdes tends to give the highest quality assemblies but is more computationally expensive and time-consuming than its competitors (van der Walt et al. 2017). The significantly longer read length and higher error rate of ONT and PacBio sequencing datasets necessitate different assembly algorithms. Long-read assemblers include Canu (Koren et al. 2017), HGAP (Chin et al. 2013), miniasm (Li 2016), MaSuRCA (Zimin et al. 2013), and Flye (Kolmogorov et al. 2019). The innovative, repeat graph approach employed by Flye performs well relative to its competitors and is rapidly becoming a tool of choice for the field (Kolmogorov et al. 2019). As mentioned above, long-read-only assemblies (particularly from ONT) have traditionally had higher error rates and benefit from a complementary Illumina dataset (although the latest ONT technology can produce accurate assemblies of microbial taxa on its own, as mentioned above). For datasets where both long-read and short-read sequencing data is available, Unicycler (Wick et al. 2017), Trycycler (Wick et al. 2021), and hybridSPAdes (Antipov et al. 2016) are available hybrid assembly tools; however, these were all developed for isolate (i.e. not metagenomic) sequencing. Draft assemblies can also be polished to further remove errors using long reads via tools including nanopolish (Loman et al. 2015) and medaka (https://github.com/nanoporetech/medaka), and/or with short reads via tools including racon (Vaser et al. 2017), pilon (Walker et al. 2014), and polypolish (Wick and Holt 2022). Polypolish was a particularly helpful advance, greatly improving short-read-based polishing in repeat and highly conserved regions, such as the rRNA genes (Baker 2022). The combination of these bioinformatics tools with third-generation long-read sequencing technologies has made it relatively easy and inexpensive to obtain accurate and complete genomes, enabling researchers to monitor reference strains for mutations and study genome-wide evolution, physiology, and pathogenesis in novel clinical and environmental isolates.

Metagenomics

Metagenomics is the study of DNA recovered directly from environmental or clinical samples, thereby containing multiple taxa (i.e. multiple genomes), which of course includes microbiome analysis. In-depth recommendations for the design and execution of a microbiome study have been expertly provided (Knight et al. 2018). Metagenomics data can be analyzed to get diversity metrics and abundance information on the taxa present. This can be done on unassembled reads using tools like MetaPhlAn4 (https://huttenhower.sph.harvard.edu/metaphlan/), [based on marker genes and part of the bioBakery suite of tools (Beghini et al. 2021)] and the Kraken family of tools [based on k-mers (Lu et al. 2022)]. In addition to taxonomic abundance information, tools such as HumanN3 (also a BioBakery tool) (Beghini et al. 2021) can obtain information regarding the metabolic pathways present in a microbiome sample, enabling analyses such as contributional diversity. This provides a significant advantage over 16S sequencing, where the functional metagenomics are not directly examined and may only be inferred linking a 16S sequence to a reference genome in a database [using a tool such as PICRUSt2 (Douglas et al. 2020)]. A species may have one 16S rRNA sequence but a significant amount of strain-to-strain intraspecies functional diversity, which will be missed in any 16S sequencing analysis. A disadvantage to most methods analyzing unassembled metagenomic reads is dependency on databases, where novel taxa or functions are likely to end up in an “unknown” bucket, which is routinely discarded by investigators (although this issue continues to decrease substantially with each subsequent version of the tools and databases).

Beyond the data generated by the unassembled reads, metagenomic datasets can be assembled to produce metagenome-assembled genomes (MAGs). A recent review covers these principles and methods in greater depth (Goussarov et al. 2022). Most of the aforementioned assembly algorithms now have versions specifically designed to handle metagenomic read sets, with metaSPAdes (Nurk et al. 2017) and MEGAHIT (Li et al. 2015) being the most commonly employed for short reads and metaFlye (Kolmogorov et al. 2020) and strainFlye (Fedarko et al. 2022) becoming the standard for long reads. Following assembly, a problem inherent with metagenomic datasets is not knowing which assembled contigs go together to form a given genome. Binning is the process of solving this problem, placing metagenomic contigs into discrete draft genomes, or “bins,” and binning typically utilizes data like k-mer frequency, GC content, and coverage, and/or alignment to references to do so. Many tools are available to perform binning or short-read-based assemblies, and several of the mainstream binning programs include MaxBin2 (Wu et al. 2016), Concoct (Alneberg et al. 2014), and MetaBat2 (Kang et al. 2019), along with high-performance newcomers such as SemiBin2 (Pan et al. 2023), Binny (Hickl et al. 2022), and MetaDecoder (Liu et al. 2022a). Recently, strategies for binning that leverage the methylation data provided by third-generation sequencing methods have been reported (Wilbanks et al. 2022). Different binning algorithms appear to produce better bins in different datasets, and indeed, tools combining composite and/or iterative binning strategies are available, including DAStool (Sieber et al. 2018), MetaWRAP (Uritskiy et al. 2018), and VEBA (Espinoza and Dupont 2022). Manual bin inspection and refinement should be performed, where possible, and have been made much easier by the Anvi’o suite of microbiome analysis programs (Chen et al. 2020a, Eren et al. 2021). There are far fewer tools to perform binning on long-read datasets, with SemiBin2 (has algorithms for both short- and long-read binning) (Pan et al. 2023) and LRBinner being the most comprehensive and recently developed (Wickramarachchi and Lin 2022). However, contigs in long-read assemblies are so much longer, and draft genomes so much more contiguous, that manual binning is much more feasible. In fact, circular (and therefore complete) chromosomes are routinely obtained using long-read metagenomic sequencing, and of course, these do not need to be binned. The ability to obtain complete and accurate genomes from metagenomic samples represents a major advance and has only become possible in a high-throughput fashion following the development of long-read sequencing (Chen et al. 2020b, Moss et al. 2020, Cusco et al. 2021, Sereika et al. 2022).

Phylogenomics

Phylogenomics is the practice of inferring evolutionary history and relatedness between different taxa and can be done using a number of different strategies. DNA sequences, including whole genome alignment, can be used and may be useful when studying the evolution of gene regulation or when reconstructing evolutionary relationships over shorter time scales. However, the use of amino acid sequences is more widely used, as they are more directly affected by natural selection, less influenced by processes such as gene duplication and horizontal gene transfer, and evolve more slowly, making it easier to reconstruct evolutionary relationships over longer time scales. PhyloSift (Darling et al. 2014), PhyloPhlAn3 (Asnicar et al. 2020), and Anvi’o (Eren et al. 2021) are widely used pipelines for performing microbial phylogenomics. These pipelines are underpinned by sequence alignment tools, such as muscle (Edgar 2004), mafft (Nakamura et al. 2018), and famsa (Deorowicz et al. 2016), as well as phylogenetic inference software, such as RAxML (Stamatakis 2014), FastTree (Price et al. 2009), and IQ-Tree (Nguyen et al. 2015). PhyloPhlAn3 [part of BioBakery3 (Beghini et al. 2021)] can easily provide taxonomic assignment to newly assembled MAGs and can perform phylogenomic analysis scalable from strain-level analysis using clade-specific markers to widely disparate clades such as whole gut microbiome phylogenomic analysis. Because phylogenomics depends, in many cases, on the alignment of widely conserved homologous core genes, it inevitably intersects with pangenomics, which is needed to identify these genes. Ideally, the lowest number of genes that still allows accurate differentiation between each taxon in the analysis should be used to reduce the computational expense of the phylogenetic inference software. The most frequent use of phylogenomics in oral microbiome research is determining the species-level taxa of a newly assembled genome or MAG. It is important to note that the concept of “species” in bacteria is not one with universally accepted traits. For the sake of ease when dealing examining massive numbers of MAGs, 95% average nucleotide identity (ANI) is the cutoff used to estimate the species level, which has been adopted by the field; however, this cutoff is not absolute and remains controversial (Jain et al. 2018, Murray et al. 2021).

Pangenomics

Pangenomics is the analysis of pangenomes, which are the collections of genes across multiple genomes. Pangenomics analysis typically identifies orthologous genes across a set of genomes and provides a list of core genes (genes present in every genome or 90%–100% of the genomes in the analysis), cloud genes (found in only a minority of genomes in the analysis), and shell genes (found in many but not all of the genomes, e.g. less than core genes but more than cloud genes); however, there are no universally accepted thresholds to determine these groups. Pangenomics is especially useful for tracing horizontal gene transfer and the evolution of specific gene clusters, including pathogenicity islands and antimicrobial resistance genes. Tools, such as Roary (Page et al. 2015), PanPhlAn3 (Beghini et al. 2021), panOCT (Fouts et al. 2012), and Anvi’o (Eren et al. 2021), have allowed pangenomics analysis at an exceptional scale and resolution. A pangenome can be parsed to identify optimal genes for phylogenetic analysis of a given dataset. These would typically be single-copy core genes that also that have maximum sequence differences across orthologs in the pangenome (so that as few genomes are identical or have a flat line in the resulting tree), but also have minimal gaps in the alignment (because phylogenetic analysis tools struggle with where to place gaps in the alignment) (described in detail at anvio.org). This type of approach will yield a bespoke phylogenetic analysis that will maximize the phylogenetic data obtained while minimizing the computational resources used and time required to perform the analysis.

Transcriptomics

Transcriptomics is the study of gene expression via sequencing of RNA and may be performed on isolates of a given taxon or multispecies samples (i.e. a metatranscriptome). For short-read-based RNAseq, gene quantification can be performed by either mapping reads to an annotated reference genome (or genomes, in the case of a metatranscriptome) or mapping reads to an annotated de novo assembly of the transcriptome (useful when reference genomes are lacking). Commonly used mapping tools for short reads include BWA-MEM (Li 2014), Bowtie2 (Langmead and Salzberg 2012), and minimap2 (Li 2018), while minimap2 can also map long reads. Common transcriptome assemblers include Trinity (Grabherr et al. 2011), RockHopper2 (Tjaden 2015), and rnaSPAdes (Bushmanova et al. 2019). Once mapped, the number of reads mapping to genes and other features can be analyzed using featureCounts (Liao et al. 2014) or a similar tool. As described in the section on Sequencing Technologies, major recent advancements to transcriptomics have come in the form of long-read RNA sequencing, the ability to detect RNA modifications and noncanonical bases, and single-cell RNAseq (scRNAseq). At this time, the application of these technologies to bacteria remains an area of active development. Current out-of-the-box RNA library preparation protocols for ONT require polyA-tailed RNA as input (eukaryotic mRNA has polyA tails but prokaryotic mRNA does not); therefore, polyA tails must be added in addition to the recommended depletion of rRNAs. Several research groups have pioneered using ONT technology for bacterial RNA-seq, and their publications provide protocols on how to do so (Pitt et al. 2020, Baker et al. 2022, Grunberger et al. 2022). Tools used to detect DNA and RNA modifications and noncanonical bases in ONT-based transcriptomics include Tombo (Oxford Nanopore Technologies, Inc.), MetaCompore, EpiNano, and MasterofPores, which have been recently benchmarked and reviewed (Wang et al. 2021, White and Hesselberth 2022).

In addition to the many free and open-source sequencing-based bioinformatics tools mentioned above, it is worth mentioning that there are also several comprehensive software suites available from vendors, such as Geneious Prime (Dotmatics, Inc.) and CLC Genomics Workbench (Qiagen, Inc.), that can do many types of the above sequencing analyses in a user-friendly graphical user interface (GUI) format (in many cases using the aforementioned individual bioinformatics tools “under the hood”), which may benefit end users with limited experience with Linux/command line-based computing skills.

The impact of sequencing-based omics on oral microbiome research

The sequencing-based omics approaches detailed above have had an extraordinary impact on our understanding of the oral microbiome. Complete genomes of oral taxa are being published at an ever-accelerating rate, making databases such as NCBI and HOMD even more useful to researchers and allowing for in-depth and accurate downstream phylogenomics and pangenomics. A number of studies have now described the oral microbiome in the context of dental caries and/or periodontal disease using shotgun metagenomics (Belda-Ferre et al. 2012, Shi et al. 2015, Yost et al. 2015, Belstrom et al. 2017, Al-Hebshi et al. 2019, Baker et al. 2021). Furthermore, several recent studies have released large numbers of oral MAGs into the public domain (Escapa et al. 2018, Pasolli et al. 2019, Baker et al. 2021, Zhu et al. 2022). While the MAGs in these large-scale, short-read-based studies are draft genomes, they represent significant progress toward identifying all of the taxa within the microbiome, as the largest study allowed mapping of ∼95% of all oral microbiome reads to the draft genomes, with only <5% of the reads being unmapped and coming from an unknown bacterial genome (Zhu et al. 2022). Crucially, between 30% and 77% of the species identified in these studies had no genomes in public repositories, illustrating that our understanding of the oral microbiota is still limited and thousands of novel taxa are still awaiting study and naming (Pasolli et al. 2019, Baker et al. 2021, Zhu et al. 2022). It is likely that many of these unknown taxa have been observed and perhaps even given a designation at the 16S level. Unfortunately, the 16S rRNA gene, due to the highly conserved elements, is only very rarely recovered in MAGs derived using short-read sequencing. Long-read metagenomic sequencing will be useful to link MAGs of novel species with their respective 16S sequences, allowing for previous 16S-based data to be leveraged for additional functional and taxonomic insight, with fewer data ending up in the “unknown taxa” bucket. Long-read-based metagenomics of the oral microbiome has been limited, but the studies that have used it were highly successful in identifying novel oral phages and examining phage pangenomics (Yahara et al. 2021), as well as obtaining complete genomes straight from saliva (Baker 2021, Baker 2022).

As these new oral genomes become available, phylogenomics analyses have identified many new species and have led to the several major phylogenetic reorganizations of taxa in the oral microbiome. Most prominent was perhaps the 2020 reorganization of the family, Lactobacillaceae (Zheng et al. 2020). This effort reclassified over 300 species in 7 genera and 2 families into one family Lactobacillaceae, which contains 31 genera, including 23 new genera that were all formerly classified as the genus Lactobacillus. The reclassification was only possible after high-quality genome sequences became available for all the type strains, as the 16S sequences were inadequate to illustrate the real phylogenetic relationships (Zheng et al. 2020). Similarly, the phylum Actinobacteria was re-classified in 2018 to include 2 orders, 10 families, and 17 genera, with over 100 species within the phylum being moved into a different genus (Nouioui et al. 2018). Diverse phylogeny within Saccharibacteria, a candidate phylum within the candidate phyla radation (CPR), continues to be resolved as new genomes become available to augment earlier 16S-based analysis (Cross et al. 2019, McLean et al. 2020, Shaiber et al. 2020, Baker 2021). On a smaller scale, phylogenomics has resolved the phylogeny of novel species within important oral taxa such as S. dentisani (Camelo-Castillo et al. 2014), Candidatus Bacteroides periocalifornicus (Torres et al. 2019), Tannerella serpentiformis (Ansbro et al. 2020), and novel taxa within Actinobacteridae (Treerat et al. 2022).

Linked closely with phylogenomics is pangenomics, and there has been no shortage of pangenome studies of oral taxa in recent years. A highlight of early pangenomics of oral bacteria was the analysis of 57 S. mutans strains to gain insight on the links phylogeny and phenotypic/virulence traits (Cornejo et al. 2013, Palmer et al. 2013). More recent work reported a detailed, updated pangenome across 244 near-complete genomes of S. mutans (Baker et al. 2022). Additional contemporary comparative genomics of S. mutans and S. sobriunus indicated a lack of phylogeographic differentiation for S. mutans but some for S. sobrinus (Achtman and Zhou 2020). Another recent study used an S. mutans pangenome to examine CRISPR spacers (Walker and Shields 2022). Beyond S. mutans, several recent studies have analyzed other Streptococcus pangenomes. A pangenome of 113 genomes from 10 Streptococcus species was utilized to gain insight into ammonia production via the arginine deiminase system and identified significant intra-species phenotypic heterogeneity (Velsko et al. 2018). Site tropism of streptococci in the oral microbiome was examined using an approach that leveraged phylogenetic and pangenomic analysis, illustrating that even closely related species such as S. mitis, S. oralis, and S. infantis specialized in different sites within the oral cavity (McLean et al. 2022). There was also substantial overlap in the core genomes of these 3 species, indicating that site-specialization is likely determined by subtle differences across the pangenome (McLean et al. 2022). Other pangenome studies examined S. intermedius and its relationship to virulence at various body sites (Sinha et al. 2021), identified homologs of adhesion and immune evasion across endocarditis and oral isolates of S. sanguinis and S. gordonii (Iversen et al. 2020), identified genomic factors influencing defense from phage and mobile genetic elements in Dolosigranulum pigrum (Flores Ramos et al. 2021), and discovered that carbohydrate utilization pathways are well-conserved across Veillonella (Mashima et al. 2021). Pangenome-based approaches also identified candidate genes involved in oral niche habitat adaptation for Rothia mucilaginosa and Haemophilus parainfluenzae (Utter et al. 2020), and illustrated niche partitioning and vast differences in metabolic repertoires between clades of oral Saccharibacteria (Shaiber et al. 2020, Baker 2021, Baker et al. 2021).

Dozens of studies have utilized transcriptomics (i.e. RNAseq) to study both individual oral bacteria under various conditions as well as communities and the entire microbiome. Early analysis of the oral metatranscriptome was provided through several studies examining both caries (Peterson et al. 2014, Do et al. 2015) and periodontal disease (Duran-Pinedo et al. 2014, Jorth et al. 2014, Yost et al. 2015, Belstrom et al. 2017, Nowicki et al. 2018), illustrating changes in both the taxonomy and functional expression in the microbiome in health versus disease. These findings were summarized in a recent review (Duran-Pinedo 2021). Metatranscriptome changes following scaling and root planning as treatment for periodontal disease were examined, showing that there was a significant effect on progressing sites but not so much in stable and fluctuating sites (Duran-Pinedo et al. 2022). Transcriptomics was used to examine the relationship between the epibiont Saccharibacteria, Nanosynbacter lyticus, and its host, Schaalia odontolytica (Hendrickson et al. 2022). A transcriptomic time course of an in vitro dental plaque biofilm maturation provided insight of transcriptional inflection points in the community associated with pH drops and blooms of acidophilic taxa such as Limosilactobacillus fermentum (Edlund et al. 2018). Recent work has illustrated the transcriptome in periodontitis in a nonhuman primate model, which supported a significant role of the adaptive immune response in the kinetics of periodontal disease progression and that aging effects on the repertoire of immunoglobulin genes are likely to contribute to an increased prevalence and severity of periodontal disease with age (Gonzalez et al. 2022). Furthermore, that the same bacterial taxa interface with host immunology differently at a healthy site compared to a diseased site (Ebersole et al. 2021). Other recent work explored the role of health-associated oral bacteria on the transcriptome of oral squamous cell carcinoma cell lines (Baraniya et al. 2022). As the oral microbiology field begins to adopt third-generation RNA sequencing, a wealth of data regarding transcriptional isoforms and RNA modification will soon become available. Several additional studies using sequencing-based omics as part of multi-omics are discussed in the Multi-omics section below.

MS-based omics

In addition to all the advances described above, which are dependent on nucleic acid sequencing, there have also been major improvements to MS-based omics analyses over the last decade. Recent innovations have made MS analyses significantly more sensitive, accurate, high-throughput, and able to detect a wider range of molecules. These have occurred via advancements at every stage of the analysis pipeline: sample preparation, ionization, separation, mass detection, and data analysis (Shuken 2023). Readers are pointed to an in-depth recent review for biological MS, broadly (Pade et al. 2023). MS-based analyses are typically either untargeted, measuring all possible analytes detectable with the given workflow, or targeted, where analysis is tailored to molecules with specific characteristics such as molecular weight and charge. Simultaneous quantitation and discovery (SQUAD) analysis, recently developed at Thermo Fisher Scientific, combines both targeted and untargeted workflows into a single-injection protocol, combining the strengths of each approach (Amer et al. 2023).

Imaging mass spectrometry (IMS) includes techniques such as matrix assisted laser desorption/ionization (MALDI), time-of-flight secondary ion mass spectrometry (TOF-SIMS), and electrospray-based desorption (DESI), which are utilized to visualize the spatial distribution and biogeography of analytes, and is at this point a mature field worthy of its own review (Chen et al. 2020a). MALDI imaging, in particular, has been widely applied to rapid clinical and diagnostic microbiology (Croxatto et al. 2012, Jang and Kim 2018). Tools such as PySM (Palmer et al. 2017), MSiReader (Nurk et al. 2017), and Ili (Protsyuk et al. 2018) have been developed to analyze and visualize IMS data.

Like sequencing-based omics, where both individual open-source tools and comprehensive vendor software suites are available, MS-based omics has individual tools as well as comprehensive software suites from vendors are options, with prime examples being Proteome Discoverer, Compound Discoverer, and Lipid Search from Thermo Fisher Scientific, and Proteoscape and Metaboscape from Brucker. Although proteomics, metabolomics, and lipidomics represent a more complete and “current” state of a given sample (i.e. rather than what is encoded for by DNA or soon-to-be translated RNA), there are unique challenges facing these omics approaches.

Proteomics

Proteins are typically higher molecular weight and more complex than metabolomic or lipidomic analytes; however, the relative wealth of proteome database data via translated RNA sequences, combined with the fact that proteins themselves are “sequences” from a finite pool of amino acids, makes proteomics datasets somewhat easier to annotate than untargeted metabolomics and lipidomics datasets. The current state of the proteomics field, including current approaches and challenges was excellently summarized recently (Shuken 2023), and bioinformatics tools for proteomics, recently comprehensively reviewed (Chen et al. 2020a). Briefly, proteomics is typically conducted with either a “bottom up” approach, which breaks proteins down into peptides prior to MS analysis, or a “top down” approach, which analyzes whole, native proteins to detect discrete proteoforms and chemical modifications (Donnelly et al. 2019). Peptide sequences are either queried in a database using an algorithm such as the workhorses SEQUEST (Eng et al. 1994) or Andromeda (Tyanova et al. 2016) or sequenced de novo using tools such as UniNovo (Jeong et al. 2013) or DeepNovo (Tran et al. 2017). Proteomics approaches can also be divided into data dependent analysis (DDA) and data-independent analysis (DIA) methods. DDA selects the most abundant peptides in each peak of the MS1 scan for the MS2 scan. Meanwhile, in DIA, ions are continuously collected and fragmented by collecting MS2 scans in overlapping m/z windows, thereby producing a complete record of all peptides in a sample (Xin et al. 2022). PEAKS Studio is a software that leverages all three techniques (peptide search, spectral library search, and de novo sequencing) (Xin et al. 2022). Proteins can be quantified using the area under the MS1 chromatogram (i.e. label-free quantification [LFQ]), which is somewhat problematic due to the compositional nature of sample-to-sample MS data, as discussed below. Alternatively, various labeling techniques such as stable isolate labeling by amino acids in cell culture (SILAC) or tandem mass tags (TMT) improve quantitative sample-to-sample reproducibility and increase throughput by allowing for multiplexing (Shuken 2023). In addition to the vendor suites of software, the MaxQuant (Wichmann et al. 2019) family of tools and MZmine 3 (Schmid et al. 2023) are free and open-source software able to perform various proteomics quantification workflows (Wichmann et al. 2019). Going forward, the application of machine learning and will further improve the sensitivity and dynamic range of proteomics via the implementation of deep learning-based spectral prediction and spectrum-centric DIA analysis (Zeng et al. 2022, Cox 2023, Neely et al. 2023).

Metabolomics and lipidomics

Several recent reviews have provided metabolomics best practices guidelines (Alseekh et al. 2021) and summarized lipidomics informatics (Ni et al. 2022), metabolomics/lipidomics separation methods (Harrieder et al. 2022), metabolite discovery (Giera et al. 2022), and the specific application of metabolomics in microbiome data (Bauermeister et al. 2022) in more depth than is provided here. In addition to vendor-specific tools such as Compound Discoverer, Lipid Search, and Metaboscape, a wealth of alternative tools (many of which are free and/or open source) are available for metabolomics analysis. These include platforms such as MetaboAnalyst 5.0 (Pang et al. 2022) and XCMS Online (Forsberg et al. 2018), which are web-based GUIs, as well as MZmine3 (Schmid et al. 2023), an open source tool to examine raw spectral files and perform custom downstream analysis, MS-DIAL (Tsugawa et al. 2020), and OpenMS (Rost et al. 2016). Unlike proteomics, metabolomics, and lipidomics data do not generate “sequences,” with the molecules being detected occupying a comparatively unlimited chemical space. Furthermore, many of the databases used for dereplication (i.e. identification of known compounds) are not freely available. As a result, a much higher percentage of the features detected in metabolomics and lipidomics datasets are unknown, with annotation rates <10% routine (de Jonge et al. 2022). In silico analyses such as molecular networking and machine learning-based annotation have been instrumental in beginning to address this challenge (de Jonge et al. 2022). Molecular networking is a visualization of spectral alignment and correlation, which enables the prediction of the chemical structure of unknown features. One such landmark tool is the Global Natural Products Social Molecular Networking (GNPS), first published in 2016, created an open-access knowledge base for organizations and enabled the sharing of MS data, which is reanalyzed as the database grows, leveraging molecular networking to help identify novel spectra (Wang et al. 2016). The GNPS led to the development of a host of integrated analysis tools to further improve annotation and analysis. The first iteration of the GNPS utilized MS-MS data exclusively, while a subsequent improvement deploys “feature-based molecular networking,” an approach that combines quantitative chromatographic peak data with qualitative MS/MS data (Nothias et al. 2020). Originally developed for liquid chromatography MS (LC-MS), the GNPS was also recently updated to enable the analysis of gas chromatography MS (GC-MS), which expands its utility to many GC-MS-based lipidomics and metabolomics analyses (Aksenov et al. 2021). Other recent innovations to MS-based omics include the use of metadata to enhance annotation of metabolomics (Gauglitz et al. 2022), native spray metal metabolomics to identify novel siderophores and other metal-binding compounds (Aron et al. 2022), and ion identity molecular networking (IIMN) to integrate chromatographic peak shape into molecular networking, enhancing annotation with molecular networks (Schmid et al. 2021). MS2Query is another recently developed tool that utilizes machine learning to identify potential analogs to unknown spectra (de Jonge et al. 2023). SIRIUS5 (Duhrkop et al. 2019) predicts the chemical formula and molecular structure of query compounds, while Qemistree is a data exploration strategy using hierarchical organization of molecular fingerprints to visualize molecular relationships as a tree, enabling the use of many further analysis tools originally designed to analyze and visualize the relatedness of DNA, such as QIIME2 (Tripathi et al. 2021). Efforts to standardize MS data and databases, such as PeakForest (Paulhe et al. 2022) and ChemFONT (Wishart et al. 2023), seek to address that major issues of data reporting and reproducibility facing the MS field (Alseekh et al. 2021). Finally, MASST is a search tool that enables uses to query spectra against all small molecule tandem-MS data in public repositories, similar to how users can query NCBI-BLAST for the source of DNA sequences (Wang et al. 2020). Going forward, these advancements in MS analysis methods are poised to increase the scale and pace of discovery in the oral microbiota and lead to novel approaches to benefit human oral health.

Impact of MS-based omics on oral microbiome research

Proteomics was utilized to study stress responses of the caries pathogen S. mutans as early as 2004 (Len et al. 2004), and many other studies have examined single oral taxa using proteomics and metabolomics. A recent study examined the S. mutans proteome during acid and oxidative stress, illustrating modules of co-expressed proteins under various stress conditions (Tinder et al. 2022). A landmark metaproteomics study of the oral microbiome identified potential biomarkers for caries (Belda-Ferre et al. 2015). Beyond the strictly microbial constituents of the oral microbiota, saliva has great diagnostic potential due to its accessibility and the large number biomarkers that can be measured using proteomics and/or metabolomics (Dawes and Wong 2019). Along those lines, the Human Salivary Proteome Wiki was recently developed and serves as a public data platform for researching and retrieving custom-curated data knowledge of the salivary proteome (Lau et al. 2021). Although lipidomics of single species, such as S. mutans (Fozo and Quivey 2004), have been performed and used to study physiology, the lipidome of the oral microbiota as a community is in need of further study. Several studies that have used MS-based omics in oral microbiome research are also mentioned in the multi-omics section below.

Compositional analysis, single-cell omics, and multi-omics

A note on compositional data and analysis tools

Nearly all omics data is compositional in nature, meaning that it is a quantitative description of parts of some whole, therefore conveying relative information. The limitations of compositional data have been excellently reviewed (Gloor et al. 2017, Morton et al. 2017, Knight et al. 2018, Morton et al. 2019b), and it is imperative that researchers are aware that omics data is compositional, perform analysis using tools designed to handle compositional data, and be cognizant of the limitations inherent to compositional data. Determining correlation is particularly intractable with compositional data, with conventional methods producing unacceptably high false discovery rates. Numerous approaches have been developed to address these problems, including ALDEx2 (Fernandes et al. 2014), ANCOM (Mandal et al. 2015), and Songbird (Morton et al. 2019b); however, none are ‘perfect’. Ultimately, it is generally best to analyze compositional data using multiple approaches and take all results with a grain of salt when forming hypotheses.

Single-cell omics

Single-cell analysis is a transformational technology, allowing for the omics analysis of individual cells and the identification of discrete biological dynamics that are obscured by the averages obtained by traditional bulk analysis. Single-cell proteomics, lipidomics, and metabolomics, based on MS, is an advancing field; however, it is still at a nascent stage even for eukaryotes and therefore will not be discussed (Couvillion et al. 2019, Perkel 2021, Tajik et al. 2022). Meanwhile, driven by advancements in microfluidics, sample handling, labeling, imaging, bioinformatics, computational biology, and machine learning, companies like 10X Genomics and Standard Biotools are making single-cell analysis of eukaryotic genomes and transcriptomes (scRNA-seq) commonplace. Challenges facing scRNA-seq in bacteria include low content of mRNA, lack of a polyA tail on mRNAs, diverse cell walls, and small size hindering microfluidic single-cell isolation (Kuchina et al. 2021). Early attempts at bacterial scRNA-seq involved using fluorescence-activated cell sorting (FACS) to distribute individual cells to wells in 96-well plates; however, this technique is low throughput, with a very high cost to examine only several hundred bacterial cells (Imdahl et al. 2020). Two concurrently developed, yet technically similar, approaches to deal with these issues are MicroSPLiT (Kuchina et al. 2021) and PETRI-seq (Blattman et al. 2020), which do not depend on single-cell isolation. Cells are permeabilized and then labeled with several rounds of split-pool barcoding of cDNA to ensure that nearly every cell has a unique barcode prior to sequencing (Blattman et al. 2020, Kuchina et al. 2021). These approaches were able to differentiate multiple transcriptional states in Bacillus subtilis and Escherichia coli, respectively (Blattman et al. 2020, Kuchina et al. 2021). More recent approaches have modified other eukaryotic scRNA-seq protocols such as multiple annealing and dC-tailing-based quantitative single-cell RNA-seq (MATQ-seq) (Homberger et al. 2023) and made use of the 10X Genomics Chromium microfluidic device (Brennan and Rosenthal 2021) to perform bacterial scRNA-seq.

In oral microbiome research, single-cell techniques have been used to isolate cells and amplify DNA to generate single-cell amplified assembled genomes (SAGs) of Saccharibacteria (Cross et al. 2019), Chloroflexi and Chlorobi (Campbell et al. 2014), Tannerella (Beall et al. 2014), Porphyromonas (McLean et al. 2013), and Desulfovibrio and Desulfobulbus (Campbell et al. 2013), and these techniques and findings were recently reviewed (Balachandran et al. 2020). Most of these organisms were present in such low numbers in the original sample that getting a substantial portion of the respective genome sequence would have been impossible without the single-cell methods. Although at this time, no studies have leveraged single-cell technology to study oral bacteria at the transcriptional level, a recent landmark study generated an atlas of human oral mucosa cells using scRNA-seq, examining healthy individuals versus periodontitis, revealing exaggerated responsiveness of stromal cells and enhanced immune cell infiltration in periodontitis (Williams et al. 2021). A recent study used also scRNA-seq to examine the expression of periodontitis susceptibility genes in human gingival cells (Caetano et al. 2022).

Multi-omics

While integrating multiple types of omics analysis is critical for microbiome research, this type of analysis introduces several additional statistical challenges as now multiple datasets that are each compositional are now being compared. Crucially, many tools specifically developed for handling compositional data lose scale invariance when applied to multi-omics datasets (Morton et al. 2019a). mmvec, a recently developed approach for analyzing multi-omics data, uses co-occurrence probabilities rather than correlations (Morton et al. 2019a). When applied to metagenome and metabolome data, it allowed researchers to identify the most likely microbe-metabolite interactions (Morton et al. 2019a). Another tool, iNetModels2, was recently developed for interactively visualizing multi-omics data (Arif et al. 2021). A recent review also comprehensively discussed tools for proteomics-centric multi-omics analyses (Rajczewski et al. 2022).

Several examples exist of published research used multi-omics data to examine the oral microbiota in various contexts. Multi-omics analysis of an in vitro oral biofilm community following a glucose pulse revealed temporal regulation of fermentation pathways affected the pH of the culture and subsequent microecology (Edlund et al. 2015). Multi-omics of dental plaque from patients with diabetes and periodontal disease identified both proteins and lipids that were associated with disease and also showed that Lautropia mirabilis synthesizes monomethyl phosphatidylethanolamine, which is rarely produced by bacteria (Overmyer et al. 2021). Multi-omics of germ-free and specific pathogen-free mice indicated that the oral microbiota influenced the permeability of the oral epithelial barrier, vis-à-vis keratinization and cell adhesion (Long et al. 2022). The relationship between the oral microbiome and chronic sleep deprivation was examined in rats, observing both taxonomic changes in the microbiota, as well as modulation of host immunological molecules (Chen et al. 2022). Finally, a recent study examined the proteome and microbiome of diseased gingival tissue (Bao et al. 2020).

Perspectives

Omics approaches have transformed our understanding of the oral microbiome and its relationship to human health, allowing for studies with a scale and resolution unimaginable 20 years ago. The HOMD now contains genomes in addition to 16S sequences and now includes the taxa from the aerodigestive tract and the oral cavity (Escapa et al. 2018). Some of the main challenges currently facing omics-based microbiome research are standardization and deposition of data in public repositories, as well as re-analysis of old data with updated reference databases. Although repositories such as the Sequence Read Archive (SRA), RefSeq, and GenBank are highly useful and do enforce some level of standardization, journals and reviewers do not always enforce the deposition of published data into these databases. Furthermore, unified repositories and data file formats are significantly more limited (and many times are vendor-specific/proprietary) for MS data. Efforts to make public databases into “living data” will also be highly useful. For example, as more and more accurate and complete genomes get deposited into the databases used to analyze the taxonomy of sequencing reads, older raw microbiome datasets can be periodically re-analyzed, and reads representing newly identified taxa can be moved from the “unknown taxa” to the proper newly identified taxa (which may change the interpretation of the results and/or identify new data trends). This is being implemented to some extent in the SRA, with entries now having a “Taxonomy Analysis” tab included in the Run Browser (Katz et al. 2021). The same is true for MS datasets, as new reference spectra get identified and added to public databases. The GNPS already has implemented “living data” using periodic reanalysis of metabolomics data stored in its repository (Wang et al. 2016). Additionally, to help reduce some of the issues in equity and reproducibility facing the field, enforcement of the publication of all analysis tools, settings, and code used in omics-based research on public repositories such as GitHub would be helpful. Continued research of the oral microbiome using omics-based approaches is needed, especially those sampling more diverse populations and performing longitudinal analysis. The discoveries enabled by this type of research will significantly improve human health.

Conflict of interest

The author declares no conflicts of interest.