Statistics-BactMentha

On this page, you can explore graphical statistical summaries of the BactMentha database.
You can also display a tableby selecting a specific type of statistic and a taxon (or all taxa).

The BactMentha database catalogs 12544 bacteria-host protein-protein interactions (PPIs) distributed as follows:
• 11612 interactions with the human host,
• 897 interactions with the mouse,
• 35 interactions with the rat.

These interactions involve 5088 distinct host proteins (respectively 4554 for humans, 521 for mice and 13 for rats) and 2970 distinct bacterial proteins (2914 interacting with human proteins, 98 with mouse proteins, and 12 with rat proteins).

In total, 102 bacterial strains from 39 distinct families are represented in BactMentha:
• 97 strains from 38 families interact with the human host,
• 27 strains from 16 families interact with the mouse,
• 6 strains from 4 families interact with the rat.

Number of bacteria-host protein-protein interactions per host taxon

Number of full annotated, partially annotated or unannotated interactions per host taxon

Global

Human

Mouse

Rat

BactMentha can provides different types of supplementary informations about the protein-protein interactons (PPIs) that can be displayed in the "Annotation" column on the right side of the resulting tables after a search in Home page, Browse page or in Advanced Search:

• bacterial protein annotations (PA):
Some bacterial proteins have been annotated as virulence factors or effectors after perfomring BLASTp against Virulence Factor Database (VFDB) and BastionHub databases. More details about the bacterial protein annotations can be found in the documentation "Data Provenance" section.

• binding regions (BR):
Experimental binding regions of interactions have been retreived from the litterature when available for either the host protein, the pathogen protein or both. Those regions might be larger and encompass the real interface of interaction due to the methods used for their detection.

• mimicINT interfaces (MI):
The mimicINT software was used to infer interfaces of interaction on host and pathogen proteins using known templates of interaction between two domains or between a domain and a Short Linear Motif (or SLiM). More details about the mimicINT workflow can be found in the documentation "Data Provenance" section.

On this side graph, we have distinguished three types of interactions:
• Fully annotated interactions correspond to interactions for which the three possible types of annotations are present (the bacterial protein is annotated and there are both experimental and inferred regions of interaction).
• Partially annotated interactions correspond to interactions for which at least one of the three types of possible annotations (the bacterial protein is annotated OR there are both experimental OR inferred regions of interaction OR any combination of two of these annotation types).
• Unannotated interactions correspond to interactions for which none of the three types of possible annotations is present (the bacterial protein isn't annotated and there is no interaction region description).

NB: mimicINT only runs for the Human taxon. So there won't be no fully annotated interactions for the other host taxa.

Here are displayed the different proportions of annotations combinations found in annotated interactions in the whole dataset and for each host taxon.

Legend:
• ___ MimicINT interfaces (MI) prediction only.
• ___ MI + BR
• ___ Experimental Binding regions (BR) only.
• ___ BR + PA
• ___ Bacterial Protein Annotation (PA) only.
• ___ MI + PA
• ___ MI + BR + PA

Enhancing Interaction Annotations with MimicINT Interfaces (MI):

Integrating MimicINT Interfaces (MI) inference has significantly improved the annotation of protein interactions in our dataset. Specifically, MI-based predictions have led to over 10% more annotated interactions (MI alone), expanding our knowledge beyond experimentally validated regions. Additionally, when MI predictions overlap with known Binding Regions (BR), they help refine existing interaction data.

This improvement is valuable for two key reasons:
• Increased Precision: MI predictions often pinpoint interaction sites with greater accuracy than BR alone. While BR annotations highlight larger protein segments necessary for interaction, MI focuses on specific protein domains or Short Linear Motifs (SLiMs), offering a more detailed understanding of interaction mechanisms.
• Guiding Experimental Validation: The high concordance between MI predictions and BR suggests that MI-based annotations are reliable. These insights can help design new experiments to validate predicted interactions, especially when no experimental data is available yet.

By integrating MI predictions, we not only enhance interaction coverage but also provide a more precise and experimentally actionable understanding of protein interactions.

Proportions of interactions annotation types (among annotated interactions) per host taxon

Global

Human

Mouse

Rat

Number of interactions (and their proportion of bacterial protein annotations) for the 10 first bacterial taxa

Global

Human

Mouse

Rat

Bacterial Protein Annotations generation:

To better understand how bacterial proteins interact with host proteins, we annotated bacterial proteins using data from the Virulence Factor Database (VFDB) and BastionHub. These annotations were assigned through a BLASTp alignment approach, where each bacterial protein in our dataset was compared to proteins in these databases.

An annotation was assigned to a bacterial protein if it met the following criteria during the BLASTp search:
• Sequence Identity ≥ 30%: The bacterial protein needed to have at least 30% similarity to a known protein in the database.
• Alignment Coverage ≥ 75%: At least 75% of the bacterial protein sequence had to align with the reference protein.

If a bacterial protein met both of these conditions, it was assigned the corresponding annotation from the database, allowing us to categorize its potential role, such as secretion systems, immune modulation, adherence, motility, and more.

Enhancing Interaction Annotations with bacterial Protein Annotations (PA):

These functional annotations enhance the database's value by providing deeper insights into pathogen strategies, such as immune evasion and toxin production, which help better understand host-pathogen interactions. They also improve data interpretation by linking bacterial proteins to key virulence mechanisms, facilitating the identification of factors contributing to infection. Moreover, they serve as a foundation for guiding experimental research, allowing to prioritize bacterial proteins for further investigation in wet-lab studies. Additionally, these annotations enable comparative analyses across bacterial taxa and host species, helping to uncover common virulence strategies and unique adaptations.

However, these protein annotations are not uniformly distributed across all pathogens, and for some bacterial species with numerous known interactions, the number of annotated proteins remains low. This suggests that these pathogens might rely on alternative infection strategies or virulence factors that are not yet well characterized, highlighting the need for further investigation to uncover potential unknown infection models.

BactMentha Database Statistics

Select a table to display :