What is PepSpace?

PepSpace is an integrated web server for the bioactivity prediction, sequence analysis, and structural characterization of peptides with biotechnological and biomedical applications. The platform was designed to support researchers in the exploration of peptide chemical space by providing computational tools for the analysis of natural and synthetic non-modified peptides.

PepSpace integrates machine learning-based prediction models with sequence and structure analysis tools, allowing users to evaluate peptides from different perspectives, including predicted biological activity, sequence similarity, conserved amino acid patterns, and structural properties. The platform is particularly focused on peptide classes of interest for drug discovery, biotechnology, molecular biology, and microbiology.

The web server is organized into three independent but complementary modules: Bioactivity Prediction, Sequence Analysis, and Structure Analysis.

Available modules

Bioactivity Prediction

The Bioactivity Prediction Module contains machine learning-based algorithms for predicting different classes of bioactive peptides, including blood-brain barrier-penetrating peptides, cell-penetrating peptides, and quorum-sensing peptides.

This module accepts peptide sequences in FASTA format and, when supported by the selected algorithm, peptide structures in PDB format. The output provides the predicted peptide class or bioactivity profile according to the selected model, supporting the screening and prioritization of peptide candidates for further computational or experimental investigation.

Sequence Analysis

The Sequence Analysis Module allows users to compare input peptide sequences, referred to as queries, with reference amino acid sequences from selected peptide classes, including positive B3PPs, CPPs, or QSPs.

This module supports two complementary alignment strategies. First, users can perform multiple global sequence alignment, which enables the visualization of conserved amino acid patterns across peptide classes. The resulting alignments are displayed as pictogram representations, also known as sequence logos. Second, users can perform pairwise local sequence alignment using a BLAST-based approach. This analysis compares each query sequence individually against the reference peptide dataset and reports the best local match, sequence identity, coverage, maximum score, E-value, number of matches and mismatches, and whether the result satisfies the predefined cutoff criterion.

Structure Analysis

The Structure Analysis Module allows users to evaluate structural features of peptides from PDB files. This module supports the structural characterization of peptide candidates by extracting and analyzing molecular information that may be relevant to peptide function, bioactivity, and classification.

The structure-based analysis can help users investigate physicochemical and conformational properties of peptides, supporting comparative studies between query peptides and reference peptide classes available in the PepSpace platform.

Supported peptide classes

B3PPs — Blood-Brain Barrier-Penetrating Peptides

Blood-brain barrier-penetrating peptides (B3PPs) are peptides with the potential to cross or interact with the blood-brain barrier. These peptides are of particular interest in drug delivery, neuropharmacology, and the development of therapeutic strategies targeting the central nervous system.

In PepSpace, B3PPs can be analyzed using bioactivity prediction, sequence comparison, and structural analysis tools.

CPPs — Cell-Penetrating Peptides

Cell-penetrating peptides (CPPs) are peptides capable of facilitating cellular uptake. They are widely studied as molecular carriers for the intracellular delivery of therapeutic molecules, nucleic acids, proteins, nanoparticles, and other bioactive compounds.

PepSpace provides tools for the prediction and analysis of CPP-like peptides using machine learning models and peptide sequence or structure information.

QSPs — Quorum-Sensing Peptides

Quorum-sensing peptides (QSPs) are signaling peptides involved in bacterial cell-to-cell communication, especially in Gram-positive bacteria. These peptides play important roles in microbial regulation, biofilm formation, virulence, and population-level responses.

In PepSpace, QSPs can be analyzed using prediction models and sequence-based comparison tools, supporting the identification and characterization of peptides potentially involved in quorum-sensing mechanisms.

Machine learning models

PepSpace includes machine learning-based models developed for the prediction of bioactive peptide classes. These models use peptide information extracted from sequence and/or structural data to classify or score peptide candidates according to their predicted bioactivity.

The currently available prediction models include BrainPepPass for B3PP prediction, ConvBoost-CPP for CPP prediction, and QuorumPep-Pred for QSP prediction. Each model was designed to support peptide screening and prioritization. The predictions should be interpreted as computational evidence and, when possible, complemented by additional in silico analyses and experimental validation.

Partner institutions

PepSpace was developed through scientific collaboration among research groups from Brazilian and international institutions:

Universidade Federal do Oeste do Para (UFOPA)

Universidade Federal do Para (UFPA)

Ghent University

Universidad del Magdalena (UniMagdalena)

Universidade Federal do Amapa (UNIFAP)

Supported by CNPq, FINEP, and CAPES.

How to cite PepSpace

If you use PepSpace in your research, please cite the PepSpace Web Server and the specific prediction model used in your analysis.

Additional references:

de Oliveira ECL, Hirmz H, Wynendaele E, Seixas Feio JA, Moreira IMS, da Costa KS, Lima AH, De Spiegeleer B, de Sales Júnior CS. BrainPepPass: A Framework Based on Supervised Dimensionality Reduction for Predicting Blood-Brain Barrier-Penetrating Peptides. Journal of Chemical Information and Modeling. 2024;64(7):2368–2382. DOI: 10.1021/acs.jcim.3c00951

Seixas Feio JA, de Oliveira ECL, de Sales CS Júnior, da Costa KS, E Lima AHL. Investigating molecular descriptors in cell-penetrating peptides prediction with deep learning: Employing N, O, and hydrophobicity according to the Eisenberg scale. PLOS ONE. 2024;19(6):e0305253. DOI: 10.1371/journal.pone.0305253

de Oliveira ECL, Santana K, Josino L, Lima e Lima AH, de Souza de Sales Júnior C. Predicting cell-penetrating peptides using machine learning algorithms and navigating in their chemical space. Scientific Reports. 2021;11:7628. DOI: 10.1038/s41598-021-87134-w