Towards the Prokaryotic Regulation Ontology: An Ontological Model to

Infer Gene Regulation Physiology from Mechanisms in Bacteria

Citlalli Mej

ıa-Almonte

and Julio Collado-Vides

Center for Genomic Science, UNAM, Av. Universidad, Cuernavaca, Mexico

Keywords:

Formal Ontology, Domain Ontology, Gene Regulation, Bacteria.

Abstract:

Here we present a formal ontological model that explicitly represents regulatory interactions among the main

objects involved in transcriptional regulation in bacteria. These formal relations allow the inference of gene

regulation physiology from gene regulation mechanisms. The automatically instantiated classes can be used

to assist in the mechanistic interpretation of gene expression experiments done at the physiological level, such

as RNA-seq. This is the ﬁrst step to develop a more comprehensive ontology focused on prokaryotic gene

regulation. The ontology is available at https://github.com/prokaryotic-regulation-ontology

1 INTRODUCTION

Since the success shown by the Gene Ontology as

a controlled vocabulary, bio-ontologies are increas-

ingly important tools in bio-informatics. However,

little has been explored regarding formal ontological

representation in the domain of bacterial gene reg-

ulation. There are two granularity levels at which

gene regulation can be studied. At the physiological

level, transcript concentration or gene product activity

is directly measured under some condition, normally

adding or depleting certain chemicals to growth me-

dia (Burstein et al., 1965). At the mechanistic level,

the effect of speciﬁc mutations on gene expression is

studied to discover the precise regulators involved in

some system (Ptashne, 1967). At this level, the most

studied mechanisms are those of transcription initia-

tion mediated by transcription factors. These proteins

can adjust gene expression to environmental require-

ments using their two main functional domains: the

effector-binding domain that senses the environmen-

tal signal and the DNA binding domain. Transcription

factors bind to DNA in sites called transcription factor

binding sites, thereby increasing or decreasing the ac-

tivity of a promoter. Promoters are the DNA regions

where transcription of transcription units (TUs) be-

gins; TUs in turn contain one or more genes. There-

fore, the expression of a TU is regulated by regula-

tion of the promoter activity. Here, we develop an

ontological model that can infer the physiology from

https://orcid.org/0000-0002-0142-5591

https://orcid.org/0000-0001-8780-7664

mechanisms of gene regulation.

The result of transcriptome analysis are sets of

genes that are either underexpressed or overexpressed

under a given condition, including the addition of

chemicals to growth media. The observation of un-

derexpression corresponds to the observation of gene

inhibition, whereas the observation of overexpres-

sion corresponds to the observation of gene induc-

tion. This means that transcriptome analysis gives us

physiological insights, rather than mechanistic ones.

The model presented here, automatically instantiates

sets of genes that are induced or repressed by some

molecule based on the mechanisms of induction or re-

pression. The ﬁnal terms will encode both the physi-

ology and the mechanisms of gene regulation (see be-

low). Thus, this ontology can help in the mechanistic

interpretation of gene expression experiments that are

done at the physiological level, such as transcriptome

analysis.

No ontology explicitly states the aim of mod-

eling gene regulation in the obo-foundry reposi-

tory (Smith et al., 2007); whereas a search in bio-

portal (Noy et al., 2009; Noy et al., 2001) only re-

trieves the Gene Regulation Ontology (GRO) (Beis-

swanger et al., 2008). This ontology includes object

properties to deﬁne agents and patients of regulation,

but it focuses on the mechanistic description of gene

regulation and it does not distinguishes the two granu-

larity levels of gene regulation described in this paper.

Thus, here we develop an ontology to represent both

mechanisms and physiology of gene regulation, the

later inferred from the former.

Mejía-Almonte, C. and Collado-Vides, J.

Towards the Prokaryotic Regulation Ontology: An Ontological Model to Infer Gene Regulation Physiology from Mechanisms in Bacteria.

DOI: 10.5220/0008387804950499

In Proceedings of the 11th International Joint Conference on Knowledge Discovery, Knowledge Engineering and Knowledge Management (IC3K 2019), pages 495-499

ISBN: 978-989-758-382-7

495

2 DEVELOPMENT PROCESS

We are using top-down ontology development ap-

proach (Noy et al., 2001). First, we included the most

general and important entities involved in regulation

of transcription initiation: transcription factor, tran-

scription factor binding site, promoter, transcription

unit, effector, etc. Second, we included the corre-

sponding biological relations among them. Third, we

created the classes that will be automatically instan-

tiated: TF bound to the DNA and regulated system.

Fourth, we formally deﬁned these classes taking ad-

vantage of the biologically relations included in the

second step (ﬁgure 2). Lastly, most speciﬁc terms

have to be generated for each speciﬁc TF, TU, pro-

moter, etc. along with their relations. The model will

automatically classify these speciﬁc entities into the

deﬁned classes (see glycolate example). RegulonDB

can be used to instantiate the ontology with knowl-

edge about Escherichia coli K-12 (Santos-Zavaleta

et al., 2018).

We are following the OBO-foundry princi-

ples. For this, we are taking advantage of the

OBO tools ROBOT (Overton et al., 2015) and

the Ontology Development Kit (https://github.com/

INCATools/ontology-development-kit). The ﬁrst one

is mainly used to extract terms and modules from ex-

isting ontologies, while the later is designed for stan-

dardized ontology documentation and release of OBO

ontologies, taking care of quality control issues. We

are using the Basic Formal Ontology as upper-level

ontology. So far, we have reused terms from six OBO-

foundry ontologies: CHEBI, GO, MSO, NCIT, OGG,

and SO (Ashburner et al., 2000; de Matos et al., 2010;

Mungall et al., 2011; Sioutos et al., 2007; He et al.,

2014) The creation of new classes and axioms was

done using Prot

e version 5.5. (Musen et al., 2015)

3 MODEL DESCRIPTION

In this paper, classes are written in italics and object

properties are written in bold face. Hierarchy is rep-

resented as indentation of bulleted lists.

3.1 An n-ary Relation to Represent the

Central Transcriptional Regulatory

Interaction

Figure 1 depicts the main elements involved in tran-

scriptional regulation along with the relations that

exist among them. These were ontologically repre-

sented as follows. Transcription factor (TF), TF bind-

ing site (TFBS), effector, and functional conformation

classes were created. Then, an n-ary relation design

pattern was used to link these four elements (Noy and

Rector, 2004). TF bound to TFBS class was cre-

ated with four properties: has binding transcription

factor, has target TFBS, is realized in functional

conformation, and has effector (Figure 1).

3.2 A Property Chain to Infer

Regulation from Anatomy

Figure 1 also depicts how the two key relations that

distinguish physiology from mechanisms of transcrip-

tional regulation were ontologically represented. The

mechanistic level describes the direct effect that a TF

bound to a TFBS has over its cognate promoter, while

the physiological level describes the effect that the

environmental condition (in our current model repre-

sented by the effector molecule) has over the expres-

sion of genes in a transcription unit. Promoter and

transcription unit classes were created. Then tran-

scription unit was related with promoter using the

property is transcribed from, whereas promoter was

related to the class TF bound to TFBS with the prop-

erty has activity regulated by. The has expression

regulated by property was created along with the fol-

lowing rule chain expressed in functional syntax (Fig-

ure 1) (Hitzler et al., 2009):

SubObjectPropertyOf(

ObjectPropertyChain( :is transcribed from

:has activity regulated by )

:has expression regulated by

)

This rule chain represents the fact that if a TU is

transcribed from a promoter, and this promoter has its

activity regulated by a TF bound to a TFBS, then this

TF bound to a TFBS regulates the expression of the

TU.

3.3 Automatic Classiﬁcation of

Regulated Systems

At the physiological level, there are only two possi-

bilities: induction or inhibition of gene expression.

At the mechanistic level, there are four possibilities.

Transcription factors bind to their cognate TFBSs and

regulate transcription only when they are in func-

tional conformation. Induction can be achieved by

activation when the binding of the effector activates

a transcription factor that increases the expression of

a TU (active conformation of TF is holo), or by de-

repression when the binding of the effector deacti-

vates a transcription factor that decreases the expres-

sion of a TU (active conformation of TF is apo). In-

hibition can be achieved by repression when the bind-

KEOD 2019 - 11th International Conference on Knowledge Engineering and Ontology Development

496

Figure 1: An n-ary relation and a property chain to represent the central regulatory interaction.

ing of the effector activates a transcription factor that

decreases the expression of a TU (active conforma-

tion is holo), or by de-activation when the binding of

the effector deactivates a transcription factor that in-

creases the expression of a TU (active conformation

is apo) (Balderas-Mart

ınez et al., 2013). All of these

cases describe the physiological response to the ap-

pearance of the effector. The disappearance of the ef-

fector reverses the response. We will treat these cases

later.

Therefore, to automatically classify TUs that are

induced or inhibited by an effector we have have cre-

ated the following subclasses of TF bound to TFBS

(Figure 2). Equivalent class axioms are shown.

• TF bound to TFBS in apo conformation is real-

ized in functional conformation some apo func-

tional conformation of TF

– TF-glycolate active in apo has effector some

glycolate

• TF bound to TFBS in holo conformation is real-

ized in functional conformation some holo func-

tional conformation of TF

– TF-glycolate active in holo has effector some

glycolate

The classes inducible system and inhibitable sys-

tem were created with the following subclasses.

Equivalent class axioms are shown.

• System induced by activation has expression in-

creased by some transcription factor bound to

TFBS in holo conformation

– System induced by activation by glycolate has

expression increased by some TF-glycolate

active in holo

• System induced by derepression has expression

decreased by some transcription factor bound to

TFBS in apo conformation

– System induced by derepression by glyco-

late has expression decreased by some TF-

glycolate active in apo

• System inhibited by repression has expression de-

creased by some transcription factor bound to

TFBS in holo conformation

– System inhibited by repression by glycolate has

expression decreased by some TF-glycolate

active in holo

Towards the Prokaryotic Regulation Ontology: An Ontological Model to Infer Gene Regulation Physiology from Mechanisms in Bacteria

497

TF bound to TFBS

TF bound to

TFBS in holo

conformation

TF-gly

regulated system

TF bound to

TFBS in apo

conformation

TF-gly

inducible system

activation

gly

Figure 2: Deﬁned classes to infer physiology from mechanisms. The outer circles represent the most general classes and

inner ovals more speciﬁc classes. On the left, the hierarchy of the molecular complex TF-TFBS-effector classes is shown.

In the text, the most speciﬁc classes are named TF-glycolate active in holo and TF-glycolate active in apo; in the ﬁgure, the

terms were shortened as TF-gly due to space issues. These classes are automatically instantiated due to the n-ary relation

shown in Figure 1. On the right, the hierarchy of effector-induced or effector-repressed systems are shown. The terms were

shortened due to space issues: activation is short for system induced by activation, derepression is short for system induced

by derepression, deactivation is short for system inhibited by deactivation, and repression is short for system inhibited by

repression, whereas gly is short for system induced by activation by glycolate, system induced by derepression by glycolate,

system inhibited by deactivation by glycolate, and system inhibited by repression by glycolate, depending on the superclass.

These classes can be automatically instantiated due to the property chain shown in Figure 1.

• System inhibited by deactivation has expression

increased by some transcription factor bound to

TFBS in apo conformation

– System inhibited by deactivation by glyco-

late has expression increased by some TF-

glycolate active in apo

In this listing of formal deﬁnitions, we included

only examples of classes deﬁned by the speciﬁc effec-

tor glycolate. The ﬁnal ontology have to be extended

to include classes for all known effectors. We plan to

do this extension using E. coli information retrieved

from RegulonDB.

4 CONCLUSIONS

An ontological model that can automatically classify

transcription units as effector-dependent repressible

or inducible systems was developed. This adds a layer

of formal knowledge to the mechanistic representa-

tion of bacterial gene regulation included in databases

like RegulonDB.

ACKNOWLEDGEMENTS

C.M.A. is a Ph.D. student from the Programa de Doc-

torado en Ciencias Biomedicas, Universidad Nacional

Autonoma de Mexico, receives fellowship 576333

from CONACYT and received ﬁnancial aid from Pro-

grama de Apoyos para Estudios de Posgrado (PAEP)

for this conference. JCV acknowledges support by

UNAM and by NIH-NIGMS grant RO1-GM110597.

REFERENCES

Ashburner, M., Ball, C. A., Blake, J. A., Botstein, D.,

Butler, H., Cherry, J. M., Davis, A. P., Dolinski, K.,

Dwight, S. S., Eppig, J. T., et al. (2000). Gene ontol-

ogy: tool for the uniﬁcation of biology. Nature genet-

ics, 25(1):25.

Balderas-Mart

ınez, Y. I., Savageau, M., Salgado, H., P

erez-

Rueda, E., Morett, E., and Collado-Vides, J. (2013).

Transcription factors in escherichia coli prefer the

holo conformation. PloS one, 8(6):e65723.

Beisswanger, E., Lee, V., Kim, J.-J., Rebholz-Schuhmann,

D., Splendiani, A., Dameron, O., Schulz, S., Hahn, U.,

et al. (2008). Gene regulation ontology (gro): design

principles and use cases. In MIE, pages 9–14.

Burstein, C., Cohn, M., Kepes, A., and Monod, J. (1965).

Role du lactose et de ses produits metaboliques dans

KEOD 2019 - 11th International Conference on Knowledge Engineering and Ontology Development

498

l’induction de l’operon lactose chez escherichia coli.

Biochimica et Biophysica Acta (BBA)-Nucleic Acids

and Protein Synthesis, 95(4):634–639.

de Matos, P., Dekker, A., Ennis, M., Hastings, J., Haug,

K., Turner, S., and Steinbeck, C. (2010). Chebi: a

chemistry ontology and database. Journal of chemin-

formatics, 2(1):P6.

He, Y., Liu, Y., and Zhao, B. (2014). Ogg: a biological on-

tology for representing genes and genomes in speciﬁc

organisms. In ICBO, pages 13–20. Citeseer.

Hitzler, P., Kr

otzsch, M., Parsia, B., Patel-Schneider, P. F.,

and Rudolph, S. (2009). Owl 2 web ontology language

primer. W3C recommendation, 27(1):123.

Mungall, C. J., Batchelor, C., and Eilbeck, K. (2011). Evo-

lution of the sequence ontology terms and relation-

ships. Journal of biomedical informatics, 44(1):87–

93.

Musen, M. A. et al. (2015). The prot

e project: a look

back and a look forward. AI matters, 1(4):4.

Noy, N. and Rector, A. (2004). Deﬁning n-ary relations on

the semantic web: Use with individuals. W3C Work-

ing Draft, 21:102.

Noy, N. F., McGuinness, D. L., et al. (2001). Ontology

development 101: A guide to creating your ﬁrst ontol-

ogy.

Noy, N. F., Shah, N. H., Whetzel, P. L., Dai, B., Dorf, M.,

Grifﬁth, N., Jonquet, C., Rubin, D. L., Storey, M.-A.,

Chute, C. G., et al. (2009). Bioportal: ontologies and

integrated data resources at the click of a mouse. Nu-

cleic acids research, 37(suppl

2):W170–W173.

Overton, J. A., Dietze, H., Essaid, S., Osumi-Sutherland,

D., and Mungall, C. J. (2015). Robot: A command-

line tool for ontology development. In ICBO.

Ptashne, M. (1967). Speciﬁc binding of the λ phage repres-

sor to λ dna. Nature, 214(5085):232.

Santos-Zavaleta, A., Salgado, H., Gama-Castro, S.,

anchez-P

erez, M., G

omez-Romero, L., Ledezma-

Tejeida, D., Garc

ıa-Sotelo, J. S., Alquicira-

Hern

andez, K., Mu

niz-Rascado, L. J., Pe

na-Loredo,

P., et al. (2018). Regulondb v 10.5: tackling

challenges to unify classic and high throughput

knowledge of gene regulation in e. coli k-12. Nucleic

acids research, 47(D1):D212–D220.

Sioutos, N., de Coronado, S., Haber, M. W., Hartel,

F. W., Shaiu, W.-L., and Wright, L. W. (2007).

Nci thesaurus: a semantic model integrating cancer-

related clinical and molecular information. Journal of

biomedical informatics, 40(1):30–43.

Smith, B., Ashburner, M., Rosse, C., Bard, J., Bug, W.,

Ceusters, W., Goldberg, L. J., Eilbeck, K., Ireland, A.,

Mungall, C. J., et al. (2007). The obo foundry: coor-

dinated evolution of ontologies to support biomedical

data integration. Nature biotechnology, 25(11):1251.

Towards the Prokaryotic Regulation Ontology: An Ontological Model to Infer Gene Regulation Physiology from Mechanisms in Bacteria

499