Page 89 - MI-2-4

Microbes & Immunity Brachyspira pilosicoli novel outer membrane proteins

2. Materials and methods The computational framework designed to select
OMPs is detailed and schematically represented in
2.1. OM β-barrel (OMBB) protein prediction Figure 1. We employed a consensus-based computational
Given that reference genomes provide a streamlined, approach to identify OMBB proteins, where the outputs
standardized, and taxonomically diverse representation from four OMP prediction tools were considered: One
of the RefSeq collection, we selected the reference strain from OMPdb (http://aias.biol.uoa.gr/OMPdb/; accessed
32
B. pilosicoli 95/1000, a porcine isolate, for our study. on April 21, 2024), MCMBB (http://athina.biol.uoa.
39
Using genome assembly ASM14372v1, the sequences of gr/bioinformatics/mcmbb/; accessed on May 1, 2024),
40
all proteins from the B. pilosicoli 95/1000 genome were TMBETADISC-radial basis function (RBF) (http://rbf.
downloaded from the National Center for Biotechnology bioinfo.tw/~sachen/OMP.html; accessed on April 23,
Information (NCBI) (https://www.ncbi.nlm.nih.gov/). 32 2024), and TMbed (https://github.com/BernhoferM/
41
42
Peptide length, molecular weight, charge, and TMbed; accessed on April 20, 2024). Protein sequences
isoelectric point for all protein sequences were determined were searched against those in the OMPdb database
−3
using the Pepstats tool from the EMBOSS package (https:// to identify homologous proteins (with E < 1×10 ; bit
www.ebi.ac.uk/jdispatcher/seqstats/emboss_pepstats). score > 100). OMPdb is a database of integral β-barrel
33
The presence of signal peptide was determined using OMPs from Gram-negative bacteria.
SignalP 5.0 (https://services.healthtech.dtu.dk/services/ MCMBB distinguishes β-barrel OMPs from globular
SignalP-6.0/; accessed on April 20, 2024) and LipoP proteins and α-helical membrane proteins. In MCMBB,
1.0 (https://services.healthtech.dtu.dk/services/LipoP- a score > 0 indicates a higher likelihood of β-barrel
1.0/; accessed on April 17, 2024). SignalP was employed conformation, whereas a score lower than zero suggests the
to predict the presence and cleavage position of signal protein is not a β-barrel. The TMBETADISC-RBF server
peptides in the protein sequences. predicts OMPs using an RBF network and position-specific
The SignalP server generates output for each protein scoring matrix profiles. TMbed, based on embeddings
sequence in the following categories: Secretory signal from protein language models, predicts the propensity of
peptide (“Sec/SPI”), lipoprotein signal peptide (“Sec/ each residue to form TMHs, transmembrane β-strands,
SPII”), Tat signal peptide (“Tat/SPI”), Tat lipoprotein signal signal peptides, or other structural elements.
peptide (“Tat/SPII”), pilin signal peptide (“Sec/SPIII”), or Using a consensus-based approach, predictions from
the absence of any signal peptide (“Other”). 34 the aforementioned tools were used to identify potential
The LipoP server predicts lipoproteins in Gram- OMBB proteins. A final list of 42 OMBB proteins was
negative bacteria and distinguishes between lipoprotein compiled based on the number of tools predicting
signal peptides, other signal peptides, and N-terminal β-barrel architecture for each protein. These proteins
transmembrane helices (TMHs). The output is classified into were categorized based on the number of tools providing
four classes: Secretory signal peptide (“SpI”), lipoprotein positive predictions. Higher confidence was assigned to
signal peptide (“SpII”), N-terminal TMH (“TMH”), and proteins predicted as OMBBs by a greater number of tools.
cytoplasmic protein (“Cyt”). The N-terminal TMH serves
35
as an anchor, stabilizing the protein within the membrane. 2.2. Structural modeling
Therefore, LipoP was used as a secondary tool to predict Structural models of the predicted OMBB proteins
signal peptides. were generated using the AlphaFold server, powered
CELLO v.2.5 (http://cello.life.nctu.edu.tw/; accessed by AlphaFold 3 (https://alphafoldserver.com/; accessed
43
on April 19, 2024) and PSORTb 3.0 (https://www.psort. on July 21, 2024). The modeling process incorporates
36
org/psortb/; accessed on April 30, 2024) were utilized to physical and chemical constraints to accurately predict
37
predict the subcellular localization of proteins. Essential protein folding, resulting in atomic coordinates for each
proteins from B. pilosicoli were predicted by performing OMBB protein. Outputs of AlphaFold 3 include confidence
BLASTP searches against the Database of Essential Genes metrics, namely: Predicted local distance difference test,
(DEG) v15.2 (https://ngdc.cncb.ac.cn/databasecommons/ predicted aligned error, predicted template modeling
database/id/229; accessed on April 2, 2024). The DEG (pTM), and interface predicted template modeling (ipTM)
38
database is a repository of essential proteins from archaea, scores. The pTM and ipTM scores assess the accuracy of the
bacteria, and eukaryotes, and assumes that proteins overall structure. 44,45 A pTM score above 0.5 and an ipTM
essential in one organism are likely to be essential in score above 0.8 indicate highly reliable predictions. The
others. Specifically, proteins with an E < 1×10 and a bit top-ranked predictions, based on predicted local distance
−3
score >100 were considered essential. difference test scores, were selected for figure generation

Volume 2 Issue 4 (2025) 81 doi: 10.36922/MI025230050