Overview

The global microbial smORFs catalogue (GMSC) is an integrated, consistently-processed, smORFs catalogue of the microbial world, combining 63,410 publicly available metagenomes from the SPIRE database and 87,920 high-quality isolated microbial genomes from the ProGenomes2 database.

A total of 4.5 billion smORFs were predicted to build the catalogue. After removing redundancy with 100% amino acid identity, we obtained a 100AA non-redundant catalogue with 964,970,496 sequences. Further, the smORFs were clustered at 90% amino acid identity resulting in 287,926,875 90AA smORFs catalogue.

In GMSC, 100AA and 90AA refer to catalogue identity thresholds rather than peptide length. 100AA is the non-redundant catalogue after collapsing exact amino acid duplicates, while 90AA groups related smORFs into family-level clusters at 90% amino acid identity.

Citation

For more details about the GMSC, please see:

Duan, Y., Santos-Júnior, C. D., Schmidt, T. S., Fullam, A., de Almeida, B. L. S., Zhu, C., Kuhn, M., Zhao, X.-M., Bork, P. & Coelho, L. P. A catalog of small proteins from the global microbiome. Nature Communications 15, 7563 (2024). DOI:10.1038/s41467-024-51894-6

Additionally, if you use GMSC in your research, please cite the above paper.

Benefits and Features

Integration:

• GMSC is available as a web resource that displays each smORFs for browsing their integrated annotations:
  • clusters
  • taxonomy classification
  • habitat assignment
  • quality assessment
  • conserved domain annotation
  • cellular localization prediction

Main purpose of GMSC:

• Expand smORF sets of global microbiomes with comprehensive annotation
• Analysis of ecological distribution patterns across taxonomy and global habitats
• Annotate smORFs of microbial genomes or genes with the resource

Searching

Search by identifier

smORFs in the catalogue are identified with the scheme GMSC10.100AA.XXX_XXX_XXX or GMSC10.90AA.XXX_XXX_XXX. The initial GMSC10 indicates the version of the catalogue (Global Microbial smORFs Catalogue 1.0). The 100AA or 90AA indicates the amino acid identity of the catalogue. The XXX_XXX_XXX is a unique numerical identifier (starting at zero). Numbers were assigned in order of increasing number of copies. So that the greater the number, the greater number of copies of that peptide were present in the raw data.

On the 100AA Sequence page, the following information is displayed for each non-redundant smORF accession.

• Protein sequence
• Nucleotide sequence
• Taxonomic assignment
• Habitat
• Protein cluster
• Quality

On the 90AA Cluster page, the following information is displayed for each family-level cluster. The 100AA members of the cluster can be displayed by pressing the show button.

• Consensus protein sequence
• Consensus nucleotide sequence
• Taxonomic assignment
• Habitat
• Number of 100AA smORFs
• Quality

Find homologues by sequence (GMSC-mapper)

GMSC-mapper is provided as a search tool for querying sequences. Users can provide contigs or protein sequences, and it will return a set of smORFs with complete annotations that match the 90AA smORF families in GMSC.

The search will take ~15 minutes. A search ID will be provided for each query. Search IDs are of the form #-xxxx, where # is an incrementing index and xxxx is a random string.

Users can wait for results on the Mapper page or look them up later from the Home page using the search ID.

GMSC-mapper can also be downloaded and run locally; see details on the GitHub page.

Browse

Users can browse by habitats and taxonomy. For example, searching for marine will match entries such as freshwater,marine,human gut. Multiple habitats can be selected.

The results are 90AA smORF families spanning the selected habitats and taxonomy. Each row represents a family-level cluster rather than an individual non-redundant 100AA sequence.

Data acquisition

63,410 assembled metagenomes were used from the SPIRE database

87,920 high-quality microbial genomes were downloaded from the ProGenomes v2 database.

Construction of GMSC

Prediction of smORFs

• We used a modified version of Prodigal to predict ORFs >= 15 bps. The ORFs that <= 300 bps were considered smORFs.

Cluster generation

All predicted smORFs were removed redundancy with 100% amino acid identity. Then they were clustered with 90% amino acid identity and 90% coverage using Linclust. As a result, the 100AA catalogue stores non-redundant sequences, while the 90AA catalogue stores family-level clusters.

Taxonomy & Habitat annotation

• Taxonomy annotation:
  • The taxonomy of assembled contigs encoding the small proteins was annotated using MMSeqs2 against the GTDB database.
  • We recorded the taxonomy of smORFs based on the taxonomy of the contigs of metagenomes or genomes of Progenome v2 database from which the smORFs were predicted. Subsequently, we assign taxonomy for 100AA and 90AA smORFs using the lowest common ancestor (LCA), ignoring the un-assigned rank to make it more specific.
• Habitat annotation:
  • We recorded the habitats of smORFs according to their source samples using the habitat microontology introduced in the SPIRE database. We further grouped the habitats into 8 broad categories: mammal gut, anthropogenic, other-human, other-animal, aquatic, human gut, soil/plant, and other.

Conserved domain annotation

The representative sequences of 90AA smORF families were searched against the NCBI CDD database by RPS-blast. Hits with a maximum e-value of 0.01 and at least 80% coverage of the PSSM length were retained and considered significant.

Cellular localization prediction

• TMHMM2 was run on the representative sequences of 90AA smORF families.
• SignalP-5.0 was run on the representative sequences of 90AA smORF families in -org gram+, -org gram-, and -org arch modes.

Quality assessment

• Terminal check: We checked for the presence of an in-frame STOP upstream of smORFs. As a smORF predicted at the start of a contig that is not preceded by an in-frame STOP codon risks being a false positive originating from an interrupted fragment.
• Antifam: We used HMMSearch to search smORFs against the Antifam database 7.0 to avoid spurious smORFs
• RNAcode: We used RNAcode, a tool to predict the coding potential of sequences based on evolutionary signatures, to identify the coding potential of 90AA smORF families containing > 8 sequences.
• Metatranscriptomes: We downloaded 221 sets of publicly available metatranscriptome data from the NCBI database paired with the metagenomic samples we used in our catalogue. We mapped reads against the representative smORFs of 90AA families by BWA. The smORFs are considered to have transcriptional evidence only if they are mapped by reads across at least 2 samples.
• Ribo-Seq: We downloaded 142 publicly available Ribo-Seq sets from the NCBI database. We mapped reads against the representative smORFs of 90AA families by BWA. The smORFs are considered to have translation evidence only if they are mapped by reads across at least 2 samples.
• Metaproteomes: We downloaded peptide datasets from 108 metaproteome projects of the PRIDE database. We exactly matched 100AA smORFs to the identified peptides of each project. If the total k-mer coverage of peptides on a smORF is greater than 50%, then the smORF is considered translated and detected.