bioRxivGeneticsPreprint
The Drosophila Secretome
Yanhui HuiD, Md Ausrafuggaman Nahid, Jenna Kalina, Austin Veal, Mujeeb QadiriiD, Ah‐Ram KimiD, Norbert PerrimoniD
Abstract
Secreted proteins are synthesized within the cell and actively transported outside, where they contribute to development, physiology, immunity, and cell to cell communication. In Drosophila melanogaster, many secreted proteins are evolutionarily conserved, making this species a powerful model for studying human development, cancer, neurobiology, and immune regulation. To assemble the Drosophila secretome, we systematically analyzed all fly protein sequences with computational tools, integrating FlyBase, UniProt and Gene Ontology annotations, and incorporating large-scale proteomics datasets. We identified 5,280 genes encoding putative secreted proteins, assigned confidence ranks and scores based on the type and strength of supporting evidence, and generated an online resource (www.flyrnai.org/apps/fly_secretome/) for exploring these data. Comparison with the human secretome shows that 56% of Drosophila secretome genes are conserved in the human genome and of these, 81% are annotated as secreted in human. In addition, comparison with Drosophila single-cell transcriptomic data revealed that secreted proteins are more tissue-specific than other genes. Finally, we demonstrate the utility of this resource by analyzing changes in expression of genes encoding putative secreted proteins during aging based on snRNA-seq datasets from the Aging Fly Cell Atlas.
- organism: Drosophila
- organism: human
- assay: protein sequence
- assay: proteomics
- assay: snRNA-seq
- research focus: database / resource
- research focus: differential expression
- research focus: phylogenetics / evolution
- research focus: cell type annotation
- approach: statistical model
In-depth summary
Claude Opus 5.5 · xhigh effort · 8 Nov 2026 · from the full text of version 2 (21 Oct 2026) · CC BY 4.0
A tiered catalogue of 5,280 Drosophila genes encoding putative secreted proteins (1,802 high-confidence) with a web portal; these genes are less conserved, more tissue-restricted and disproportionately down-regulated with age.
Large-scale experimental studies have identified hundreds of secreted proteins in Drosophila, but no analysis had tried to catalogue all putative candidates. This resource merges FlyBase, UniProt, Gene Ontology and hemolymph proteomics with proteome-wide signal peptide, localization, transmembrane and extracellular vesicle evidence, ranking each gene by confidence tier and score in a searchable portal.
Key results
ShownThe putative fly secretome holds 5,280 protein-coding genes. 1,802 high-confidence genes come from curated annotation or experimental data. 3,478 low-confidence genes come from computational predictions (signal peptides, predicted cell-surface proteins) or Vesiclepedia. Each gene also gets an evidence-based score.
Four curated sources gave 1,802 unique genes: 1,130 (63%) from a single source and 672 (37%) from two or more. Of these, 1,569 (87%) had at least one isoform with a predicted signal peptide. Another 2,509 unannotated genes had predicted signal peptides, and Vesiclepedia added 451 genes. · Assembly of a Drosophila secretome database
ShownSecretome genes have a human ortholog less often than other fly genes, and the gap remains after stratifying by protein length. Most of the conserved secretome genes have orthologs that are annotated as human secreted genes.
DIOPT v10 (high/moderate): 56% (2,955 of 5,280) of secretome genes have a human ortholog vs 79% of non-secretome genes; 2,402 (81%) of these map to a 6881-gene human secretome reference. Proteins >200 amino acids: 69% vs 83%; 200 amino acids or shorter: 31% vs 60%. Short proteins (<200 amino acids): 28% of secreted vs 19% of other genes. · 1.) Conservation
SuggestedIn Fly Cell Atlas (FCA) pseudo-bulk data, secretome genes fall into tissue-specific categories more often than the rest of the proteome, and into 'Expressed in All' less often. Uncharacterized (CG) secretome genes are more tissue-restricted than characterized ones.
250 cell types were pooled into 14 systems as counts-per-million (CPM) pseudo-bulk and classed with the Human Protein Atlas 5-fold scheme; the text does not give proportions for secretome vs rest. Uncharacterized vs characterized secretome genes: tissue-specific 22% vs 14%, expressed in all 4% vs 24%, not detected 40% vs 18%. CG genes make up 44% of the secretome vs 40% of other genes. · 2.) Tissue expression
SuggestedIn Aging Fly Cell Atlas body data, high-confidence secreted genes change expression between 70-day and 5-day flies more often than other genes, mostly downward. In hemocytes, however, more immune genes go up than down.
Per-system comparison, adjusted p < 0.05 and |log2 fold change| > 1: 39% of high-confidence secreted genes are down-regulated in one or more tissues vs 22% of non-secreted genes, and 49% of them are differentially expressed vs 36% of all protein-coding genes. Down-regulated secreted genes outnumber up-regulated ones, often by 5–15 fold; female reproduction and gut have the most. · Global expression changes in the secretome…; Figure legends
ShownA public web portal supports single-gene and batch lookup and lets users browse genes filtered by FCA expression. It links each gene to its evidence, predictions, proteomics, orthologs, candidate receptors and RNAi stocks. Existing RNAi lines cover most secretome genes.
The portal has a 13-tab menu: bulk RNA-seq (modENCODE, FlyAtlas2), scRNA-seq pseudo-bulk, markers and differential expression, a proteomics-based secretion destination, DIOPT orthologs, PANGEA gene sets, FlyPhoneDB/FlyPredictome receptors and RNAi stocks. VDRC, TRiP and NIG RNAi lines cover 92% of secretome genes, including 85% of 2,208 uncharacterized ones; UAS-cDNA stocks cover 33% and 15% respectively. · 2.) Tissue expression; Drosophila Secretome online resource
Methods, data and reproducibility
Curated sets (FlyBase, UniProt, GO, TurboID proteomics) were merged with proteome-wide signal peptide, transmembrane, DeepLoc 2.0 and ProP predictions, published cell-surface predictions and Vesiclepedia genes, then tiered and scored. Genes were mapped to other secretomes via DIOPT, profiled in Fly Cell Atlas pseudo-bulks, and tested for aging changes in the Aging Fly Cell Atlas with Seurat FindMarkers.
- Full Drosophila proteome translation FASTA (FlyBase r6.69, FB2026_03), used as input to all prediction tools · Drosophila melanogaster
- Four curated sources: FlyBase receptor ligand gene group, UniProt 'Secreted' (SL-0243), GO extracellular region, and a TurboID ER-lumen hemolymph mass spectrometry study · Drosophila melanogaster · n = 1,802 genes combined; proteomics: 535 circulating proteins, 10 larval tissues
- Vesiclepedia extracellular vesicle proteins, Drosophila subset · Drosophila melanogaster · n = 451 genes added
- Human secretome reference assembled from Human Protein Atlas, SPD, SPRomeDB and SEPDB · human · n = 6881 genes
- Mouse and rat secretomes from SEPDB and a zebrafish secretome from a publication, used for conservation comparisons · mouse, rat, zebrafish
- Fly Cell Atlas '10x VSN All (Stringent)' single-nucleus RNA-seq of adult flies · Drosophila melanogaster · n = 580,000 nuclei; 250 cell types grouped into 14 systems
- Aging Fly Cell Atlas single-nucleus RNA-seq, body only (head excluded) · Drosophila melanogaster · n = ages 5, 30, 50, 70 days; 70 vs 5 days compared · GSE218661
- Methods & tools
- SignalP 6.0, TargetP 2.0, Phobius, TOPCONS, DeepLoc 2.0, DeepTMHMM, TMHMM 2.0, ProP 1.0, DIOPT v10, PANGEA, Seurat v5.5.0, ComplexHeatmap v2.26.1
- Data & accessions
- https://www.flyrnai.org/apps/fly_secretomehttps://www.flycellatlas.org/#datahttps://s3ftp.flybase.org/genomes/dmel/dmel_r6.69_FB2026_03/fasta/dmel-all-translation-r6.69.fasta.gzhttps://www.proteinatlas.org/download/proteinatlas.tsv.ziphttps://www.microvesicles.org/https://flybase.org/reports/FBgg0001105
- Reagents & environment
- FlyBase FB2026_03 (proteome r6.69); SignalP 6.0 (fast and slow modes combined), TargetP 2.0, DeepLoc 2.0, TMHMM 2.0 and ProP 1.0 run locally; DIOPT v10; R v4.5.0, Seurat v5.5.0, ComplexHeatmap v2.26.1; Human Protein Atlas download of June 3rd 2026. No container or requirements file is mentioned.
- Assay conditions & compute
- Portal and MySQL database are hosted on the Harvard Medical School O2 high-performance computing cluster; hardware and run time for the local prediction runs are not stated.
- Reproducibility gaps
- No analysis code or scripts are mentioned. Phobius and TOPCONS were run on web servers; the proteomics list comes from supplementary table 3B of a cited paper and the cell-surface lists from two publications; the zebrafish secretome source is described only as a publication.
Limitations (6)
- The low-confidence tier, a large share of the resource, rests on signal peptide predictions that were not experimentally validated, so false positives and negatives are expected at the tools' inherent rates. · stated by the authors
- Proteins that enter the endoplasmic reticulum but are retained intracellularly are flagged from DeepLoc2 predictions with a low score; some may be misclassified given the limits of localization prediction. · stated by the authors
- Tissue expression analyses pool single-nucleus data into broad systems by pseudo-bulk aggregation, which can mask cell-type-specific patterns. · stated by the authors
- Conservation analysis used only sequence-based orthology, so diverged signaling molecules may be missed; the authors propose structural search such as Foldseek as a remedy. · stated by the authors
- Aging differential expression used Seurat FindMarkers, which treats individual cells as observations rather than testing at the sample level; the methods describe the comparison as per cell type, the results as per system. The idea that secreted factors mediate aging is proposed but not tested here. · noted in this summary
- The main secretome vs non-secretome contrasts (conservation, tissue specificity, share of aging-related expression changes) are reported as proportions without statistical tests, and the tier and score scheme is not benchmarked against a held-out set of known secreted proteins. · noted in this summary
Read this if you need a confidence-ranked list of Drosophila secreted protein candidates with human orthology, single-nucleus expression and RNAi stock links to pick ligands or uncharacterized genes for functional screens.
- Citations
- 0
Reproducibility ledger
Nobody has reported trying to reproduce this paper yet. Did you replicate an experiment or run the code? Your report saves the next lab a week.
Sign in to file a reproducibility report.
Related papers
- Multi-omic mapping of Drosophila protein secretomes reveals tissue-specific origins and inter-organ trafficking
Justin A. Bosch, Pierre M. Jean Beltran, … Norbert Perrimon·Nature CommunicationsPeer-reviewed·Apr 20, 2026
- Research gaps and new insights in the intriguing evolution of Drosophila seminal proteins
J. Hurtado, F. C. Almeida, … E. Hasson·bioRxiv·May 11, 2021
- In-silico definition of the Drosophila melanogaster matrisome
M. N. Davis, S. Horne-Badovinac, A. Naba·bioRxiv·Aug 2, 2019
- Quantitative proteomics identification of seminal fluid proteins in male Drosophila melanogaster
Irem Sepil, Ben R. Hopkins, … Stuart Wigby·Molecular & Cellular ProteomicsPeer-reviewed·Oct 4, 2018
- Gene age shapes functional and evolutionary properties of the Drosophila seminal fluid proteome
José M. Ranz, Carolina Flacchi, … Alberto Civetta·Proceedings of the National Academy of SciencesPeer-reviewed·Oct 7, 2025
- Evidence for Transcriptomic Conservation Between the Main Cells of the Drosophila Prostate-Like Accessory Gland and Basal Cells of the Mammalian Prostate.
S. Jaimian Church, April R. Kriebel, … Laura Buttitta·bioRxiv·Jun 5, 2025