Benchmark data of EpiBuilder 2.0: CPU/GPU performance, BepiPred v3.0 vs. EpiDope v0.3 comparison, and epitope predictions for five pathogen proteomes

This record contains the raw data generated in the computational experiments of the article "EpiBuilder 2.0: A Dockerized and User-Friendly Web Platform for in silico Analysis of Linear B-cell Epitopes". Source code: https://github.com/labioinfoufsc/EpiBuilder. The data were used to (i) benchmark the processing time of the EpiBuilder 2.0 workflow on CPU and GPU on two different machines, (ii) compare the epitope predictor integrated into EpiBuilder 2.0 (BepiPred v3.0) with an alternative deep-learning predictor (EpiDope v0.3), in processing time and in per-residue concordance, and (iii) carry out the biological evaluation of the predicted linear B-cell epitopes (N-glycosylation filtering, subcellular localization, BLASTp matches against the IEDB and against other proteomes of the same genus). All runs were performed in triplicate. The archive contains only the generated outputs and the shell scripts used to produce them; the input proteomes are public UniProt reference proteomes and are not redistributed. Hardware configurations Machine A (Intel Core i9-14900K @ 6.0 GHz, 80 GB DDR5 6000 MHz): configuration A – CPU only (machine_A/cpu/); A1 – NVIDIA RTX 3080, 10 GB (machine_A/RTX_3080/); A2 – NVIDIA RTX A4500, 20 GB (machine_A/RTX_A4500/). Machine B (Intel Xeon E7-4830 v2 @ 2.20 GHz, 295 GB DDR3 1600 MHz): configuration B – CPU only (machine_B/cpu/). Input proteomes (UniProt, downloaded on 6 November 2026) Influenza B virus B/Lee/1940 – UP000008158 (11 proteins) Zaire ebolavirus – UP000007209 (9 proteins) Borrelia burgdorferi B31 – UP000001807 (1,273 valid proteins) Staphylococcus aureus – UP000008816 (2,888 valid proteins) Candida albicans – UP000000559 (5,977 valid proteins) Software and parameters BepiPred v3.0 (Docker image bioinfoufsc/bepipred3), CPU and GPU. EpiDope v0.3 (flomock/epidope:v0.3), CPU only. EpiBuilder 2.0 core workflow in Nextflow (bioinfoufsc/epibuilder-core): epitope extraction, physicochemical properties, N-glycosylation, PSORTb v3.0 / WoLF PSORT v0.2, BLASTp v2.17 against the IEDB (~155,279 B-cell epitopes) and reference proteomes of each genus. Parameters: epitope length 10–30 aa; BepiPred v3.0 threshold 0.1512; BLAST minimum identity 90% and minimum query coverage 90%; localization none (viruses), gram_neg (B. burgdorferi), gram_pos (S. aureus), fungi (C. albicans). Contents of EpiBuilder2_experiments.tar.gz scripts/ – experiment_bepipred.sh (BepiPred v3.0, CPU or GPU), experiment_epidope.sh (EpiDope v0.3), experiment.sh (one EpiBuilder 2.0 run) and experiment_triplicate.sh (loop over organisms and replicates). machine_A/ and machine_B/ – one folder per configuration and tool (bepipred, epidope, epibuilder), with one subfolder per organism (INFLUENZA, EBOLA, BORRELIA, STAPHYLOCOCCUS, CANDIDA) and replicate1–replicate3. bepipred/ – FASTA validation output, per-residue scores (raw_output.csv), execution logs and wall-clock times (execution_times.tsv). epidope/ – per-residue scores (epidope_scores.csv), predicted epitopes, per-protein tables and plots, and execution times. epibuilder/ – epitope-level results (epitope-detail.tsv, epitopes.fasta), per-protein summary, per-residue scores, subcellular localization, BLASTp hits against the IEDB and each reference proteome, consolidated spreadsheet (epibuilder.xlsx) and Nextflow trace, report, timeline and DAG (reports/). Use in the publication CPU vs. GPU processing time of the workflow (configurations A, A1, A2 and B) and comparison between machines: bepipred/*/execution_times.tsv and epibuilder/*/replicateN/reports/trace.txt. BepiPred v3.0 vs. EpiDope v0.3 processing time on CPU: execution_times.tsv of both tools. Per-residue concordance between BepiPred v3.0 and EpiDope v0.3 (Cohen's kappa, Jaccard index, Spearman's rho): raw_output.csv and epidope_scores.csv. Biological evaluation, IEDB matches and cross-species epitope sharing: epibuilder/ outputs. Supplementary Materials 1 (execution times) and 2 (complete epitope-level output) were derived from these files. A detailed description of every file, and instructions for reproducing the experiments, are provided in the README.md file of this record. Machine names were anonymised as Machine A and Machine B, as in the article.

Authors

Institutions

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-11-17
DOI
https://doi.org/10.5281/zenodo.24294894
Primary Topic
vaccines and immunoinformatics approaches
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Benchmark data of EpiBuilder 2.0: CPU/GPU performance, BepiPred v3.0 vs. EpiDope v0.3 comparison, and epitope predictions for five pathogen proteomes

Glauber Wagner, Beatriz Figueredo, Renato Simões Moreira, Bruna Caroline Russi et al.
Zenodo (CERN European Organization for Nuclear Research)
vaccines and immunoinformatics approaches
article

Benchmark data of EpiBuilder 2.0: CPU/GPU performance, BepiPred v3.0 vs. EpiDope v0.3 comparison, and epitope predictions for five pathogen proteomes

Glauber Wagner, Beatriz Figueredo, Renato Simões Moreira, Bruna Caroline Russi, Lais Apolinario Sales
article en

Abstract

This record contains the raw data generated in the computational experiments of the article "EpiBuilder 2.0: A Dockerized and User-Friendly Web Platform for in silico Analysis of Linear B-cell Epitopes". Source code: https://github.com/labioinfoufsc/EpiBuilder. The data were used to (i) benchmark the processing time of the EpiBuilder 2.0 workflow on CPU and GPU on two different machines, (ii) compare the epitope predictor integrated into EpiBuilder 2.0 (BepiPred v3.0) with an alternative deep-learning predictor (EpiDope v0.3), in processing time and in per-residue concordance, and (iii) carry out the biological evaluation of the predicted linear B-cell epitopes (N-glycosylation filtering, subcellular localization, BLASTp matches against the IEDB and against other proteomes of the same genus). All runs were performed in triplicate. The archive contains only the generated outputs and the shell scripts used to produce them; the input proteomes are public UniProt reference proteomes and are not redistributed. Hardware configurations Machine A (Intel Core i9-14900K @ 6.0 GHz, 80 GB DDR5 6000 MHz): configuration A – CPU only (machine_A/cpu/); A1 – NVIDIA RTX 3080, 10 GB (machine_A/RTX_3080/); A2 – NVIDIA RTX A4500, 20 GB (machine_A/RTX_A4500/). Machine B (Intel Xeon E7-4830 v2 @ 2.20 GHz, 295 GB DDR3 1600 MHz): configuration B – CPU only (machine_B/cpu/). Input proteomes (UniProt, downloaded on 6 November 2026) Influenza B virus B/Lee/1940 – UP000008158 (11 proteins) Zaire ebolavirus – UP000007209 (9 proteins) Borrelia burgdorferi B31 – UP000001807 (1,273 valid proteins) Staphylococcus aureus – UP000008816 (2,888 valid proteins) Candida albicans – UP000000559 (5,977 valid proteins) Software and parameters BepiPred v3.0 (Docker image bioinfoufsc/bepipred3), CPU and GPU. EpiDope v0.3 (flomock/epidope:v0.3), CPU only. EpiBuilder 2.0 core workflow in Nextflow (bioinfoufsc/epibuilder-core): epitope extraction, physicochemical properties, N-glycosylation, PSORTb v3.0 / WoLF PSORT v0.2, BLASTp v2.17 against the IEDB (~155,279 B-cell epitopes) and reference proteomes of each genus. Parameters: epitope length 10–30 aa; BepiPred v3.0 threshold 0.1512; BLAST minimum identity 90% and minimum query coverage 90%; localization none (viruses), gram_neg (B. burgdorferi), gram_pos (S. aureus), fungi (C. albicans). Contents of EpiBuilder2_experiments.tar.gz scripts/ – experiment_bepipred.sh (BepiPred v3.0, CPU or GPU), experiment_epidope.sh (EpiDope v0.3), experiment.sh (one EpiBuilder 2.0 run) and experiment_triplicate.sh (loop over organisms and replicates). machine_A/ and machine_B/ – one folder per configuration and tool (bepipred, epidope, epibuilder), with one subfolder per organism (INFLUENZA, EBOLA, BORRELIA, STAPHYLOCOCCUS, CANDIDA) and replicate1–replicate3. bepipred/ – FASTA validation output, per-residue scores (raw_output.csv), execution logs and wall-clock times (execution_times.tsv). epidope/ – per-residue scores (epidope_scores.csv), predicted epitopes, per-protein tables and plots, and execution times. epibuilder/ – epitope-level results (epitope-detail.tsv, epitopes.fasta), per-protein summary, per-residue scores, subcellular localization, BLASTp hits against the IEDB and each reference proteome, consolidated spreadsheet (epibuilder.xlsx) and Nextflow trace, report, timeline and DAG (reports/). Use in the publication CPU vs. GPU processing time of the workflow (configurations A, A1, A2 and B) and comparison between machines: bepipred/*/execution_times.tsv and epibuilder/*/replicateN/reports/trace.txt. BepiPred v3.0 vs. EpiDope v0.3 processing time on CPU: execution_times.tsv of both tools. Per-residue concordance between BepiPred v3.0 and EpiDope v0.3 (Cohen's kappa, Jaccard index, Spearman's rho): raw_output.csv and epidope_scores.csv. Biological evaluation, IEDB matches and cross-species epitope sharing: epibuilder/ outputs. Supplementary Materials 1 (execution times) and 2 (complete epitope-level output) were derived from these files. A detailed description of every file, and instructions for reproducing the experiments, are provided in the README.md file of this record. Machine names were anonymised as Machine A and Machine B, as in the article.

Zenodo (CERN European Organization for Nuclear Research)
Universidade Federal de Santa Catarina (BR), Instituto Federal de Educação, Ciência e Tecnologia de Santa Catarina (BR)
vaccines and immunoinformatics approaches
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon