Skip to content

Cookbook

These pages contain suggestions and guidelines for representing scientific findings in BEL.

  • Representation of Experimental Data
  • Statement Annotations
  • Modified Proteins
  • Reactions
  • Protein-Protein Interactions
  • Protein Families

Representation of Experimental Data

In a causal BEL Statement, the subject term frequently represents an experimentally manipulated entity while the object term represents a measured entity. Our best practices apply different levels of inference for mapping subject and object terms, particularly for representing 'omic data.

  • Subject Terms (Perturbations)
  • Relationships
  • Object Terms (Measurements)

Subject Terms (Perturbations)

  • BELv2How should I represent chemical inhibitor experiments?
  • How do I represent experiments that use site-directed mutants?
  • How do I represent observations resulting from manipulation of two or more entities?
  • How should I represent gene knock out or RNAi experiments?
  • How should I represent overexpression experiments?
  • When should I use the protein abundance vs. the activity of a protein?

BELv2 How should I represent chemical inhibitor experiments?

For experiments where protein activity is perturbed with a chemical inhibitor, we generally use the chemical as the subject term and not the activity of the target protein. In many cases, the effects of the chemical are not specific to the intended target. This representation approach avoids unintended attribution of off-target effects of a chemical to the target protein.

For example, treatment of cells with the PI3 kinase inhibitor LY294002 significantly decreases expression of TGFB2 RNA (http://www.ncbi.nlm.nih.gov/pubmed?term=20629536[PMID 20629536]):

a(SCHEM:"LY 294002") -| r(HGNC:TGFB2)

In a case where more information is available, the protein activity targeted by the inhibitor can be used as the subject term. For example, if the effect of LY 29004 on TGFB2 RNA expression was demonstrated to require the PIK3CA gene, we could represent the subject term as the kinase activity of the PIK3CA protein.

act(p(HGNC:PIK3CA), ma(kin)) -> r(HGNC:TGFB2)

How do I represent experiments that use site-directed mutants?

The artificial (laboratory) creation of sequence variants is often used to investigate the effects of protein activity or specific post-translational modifications. These include proteins altered to be constitutively active or dominant negative, as well as proteins with specific amino acid residues altered to prevent phosphorylation. While many of these sequence alterations can be precisely represented with BEL, this may not be the best approach to capturing the observations from experiments that use these constructs.

Non-phosphorylatable mutant

In this example, mutation of FOXO1 serine 256 to alanine is used to block phosphorylation at 256 (S256A), a site phosphorylated by AKT. The S256A mutation was found to impair phosphorylation of threonine 24 and serine 319 by AKT (http://www.ncbi.nlm.nih.gov/pubmed?term=11237865[PMID 11237865]). We could represent this observation as follows:

p(HGNC:FOXO1, var("p.Ser256Ala")) =| p(HGNC:FOXO1, pmod(Ph, Ser, 256))
p(HGNC:FOXO1, var("p.Ser256Ala")) =| (p(SFAM:"AKT Family") => p(HGNC:FOXO1, pmod(Ph, Thr, 24)))
p(HGNC:FOXO1, var("p.Ser256Ala")) =| (p(SFAM:"AKT Family") => p(HGNC:FOXO1, pmod(Ph, Ser, 319)))

The first statement indicates that phosphorylation at S256 is blocked by mutation of S256 to alanine. The next two statements indicate that the S256A mutation decreases phosphorylation of FOXO1 threonine 24 and serine 319 by AKT. However, we are not generally interested in the effects of a lab-created mutant like S256A so much as the role of phosphorylation at serine 256 on phosphorylation of the other two sites. Thus, we recommend the following representation:

p(HGNC:FOXO1, pmod(Ph, Ser, 256)) => (p(SFAM:"AKT Family") => p(HGNC:FOXO1, pmod(Ph, Thr, 24)))
p(HGNC:FOXO1, pmod(Ph, Ser, 256)) => (p(SFAM:"AKT Family") => p(HGNC:FOXO1, pmod(Ph, Ser, 319)))

Here, the statements indicate that phosphorylation of FOXO1 at S256 increases the phosphorylation of T24 and S319 by the kinase activity of AKT. While both representations are accurate, the second version is better suited to integrating other information about the role of FOXO1 phosphorylation at S256 into a cohesive, traversable model.

How do I represent observations resulting from manipulation of two or more entities?

In some cases an experiment has a complex perturbation, where manipulations of multiple biological entities are required for an effect. Multiple BEL abundance terms can be represented together as the subject of a BEL Statement by using the compositeAbundance() or composite() function.

In this example, TGF-beta cooperates with IL-6 to generate T-helper 17 cells (http://www.ncbi.nlm.nih.gov/pubmed?term=17918200[PMID 17918200]):

composite(p(MGI:Tgfb1), p(MGI:Il6)) -> bp(GO:"T-helper 17 cell differentiation")

If the two manipulated components are known to physically interact (such as a receptor and it's ligand), we recommend inferring their effects rather than using a composite term.

In this example, both Met and Hgf (the Met ligand) are required for increased expression of integrin Itgav RNA (http://www.ncbi.nlm.nih.gov/pubmed?term=16710476[PMID 16710476]):

Not recommended:

 composite(p(MGI:Hgf), p(MGI:Met)) -> r(MGI:Itgav)

Recommended:

kin(p(MGI:Met)) -> r(MGI:Itgav)

p(MGI:Hgf) -> r(MGI:Itgav)

Because Hgf binds to and directly activates Met, the effect of Met and Hgf together on Itgav RNA expression can be inferred to result from Met activity.

How should I represent gene knock out or RNAi experiments?

Our general practice is to represent the subject term for experiments where the perturbation is a gene deletion or RNAi knockdown as the abundance of the corresponding protein.

Gene knockouts

In this example, mice with a gene deletion of Nfe2l2 express reduced mRNA of the glutathione S transferase Gsta1 compared to wild-type mice (http://www.ncbi.nlm.nih.gov/pubmed?term=11991805[PMID 11991805]):

p(MGI:Nfe2l2) -> r(MGI:Gsta1)

RNA interference

In this example, knockdown of PTEN using RNA interference results in increased CDKN1A protein levels (http://www.ncbi.nlm.nih.gov/pubmed?term=17300726[PMID 17300726]):

p(HGNC:PTEN) -| p(HGNC:CDKN1A)

We assume that the effects of PTEN RNAi are due to knock down of PTEN protein. Decreased PTEN protein resulting in increased CDKN1A protein is interpreted as PTEN decreases CDKN1A protein.

It is generally preferable to represent the subject term as the protein abundance and not an activity of the protein, particularly for 'omic experiments. See also When should I use the protein abundance vs. the activity of a protein?

How should I represent overexpression experiments?

For experiments where the perturbation is overexpression of DNA or RNA for the purpose of overexpressing a protein, we generally represent the subject term as a protein abundance.

In this example, SIAH2 and repp86 (TPX2) proteins interact, and overexpression of SIAH2 by transfection increases degradation of TPX2 protein (http://www.ncbi.nlm.nih.gov/pubmed?term=17716627[PMID 17716627]):

p(HGNC:SIAH2) => deg(p(HGNC:TPX2))

The statement is modeled as direct because the subject and object term proteins interact.

While it would be technically correct to represent overexpressions achieved via DNA transfection as gene abundances and those from mRNA transfections as RNA abundances, this distinction is not useful for applications like Whistle and pathfinding. It is generally preferable to represent the subject term as the protein abundance and not an activity of the protein, particularly for 'omic experiments, if it is not clear that the activity is required or responsible for the effect.

When should I use the protein abundance vs. the activity of a protein?

Many experimental perturbations involve the overexpression or knockdown of protein abundance. We generally represent this type of experiment using a protein abundance as the subject term instead of an activity (e.g., kinase or phosphatase) of the protein. While many proteins have a known activity, this activity is not always responsible for all downstream effects of overexpression or knockdown; the same effects occur when either wild-type or catalytically inactive forms are expressed.

Below are examples of BEL statements for cases where:

. the effects of increased or decreased protein abundance are not due to the catalytic activity of the protein, . effects are likely due to an increase or decrease in the activity of the protein, and . not enough information is available.

Effects are not due to the catalytic activity of the protein

Example 1. A mutant telomerase (TERT) protein lacking telomerase activity retains its effects on keratinocyte proliferation (http://www.ncbi.nlm.nih.gov/pubmed/18208333[PMID 18208333]):

p(MGI:Tert) -> bp(GO:"cell proliferation")
act(p(MGI:Tert), ma(cat)) causesNoChange bp(GO:"cell proliferation")

Example 2. The serine-threonine kinase RIPK1 activates NF-kappaB through a mechanism that does not involve the protein's kinase activity (http://www.ncbi.nlm.nih.gov/pubmed/20354226[PMID 20354226]).

p(HGNC:RIPK1) -> act(p(GOCC:"NF-kappaB complex"), ma(tscript))
act(p(HGNC:RIPK1), ma(kin)) causesNoChange act(p(GOCC:"NF-kappaB complex"), ma(tscript))

Effects are likely due to the activity of the protein

Example. Genes are differentially expressed in wild-type vs. Met knock out mouse hepatocytes, only after treatment with Hgf, the Met ligand (http://www.ncbi.nlm.nih.gov/pubmed/16710476[PMID 16710476]). These genes include integrins Itgav, Itga3, and Itgb1.

act(p(MGI:Met), ma(kin)) -> r(MGI:Itgav)
act(p(MGI:Met), ma(kin)) -> r(MGI:Itga3)
act(p(MGI:Met), ma(kin)) -> r(MGI:Itgb1)

Because both Met and the Met ligand are required for the increase in gene expression, we use the activity of Met as the subject term.

Not enough information is available

Example. Genes are differentially expressed in the pancreas of mice with a pancreas-specific beta-catenin deletion (http://www.ncbi.nlm.nih.gov/pubmed/17222338[PMID 17222338]). These include decreased expression of the hedgehog interacting protein Hhip in the knock-out compared to wild-type.

p(MGI:Ctnnb1) -> r(MGI:Hhip)

In this case, not enough information is available to determine if the change in Hhip expression is due to Ctnnb1 function in transcription, cell adhesion, or another role.

Relationships

When should I use a correlative relationship?

Correlative Relationships, Correlative relationships are more appropriate than Causal Relationships, causal relationships to represent observations that do not clearly result from an experimental perturbation. BEL causal relationships include increases, decreases, directlyIncreases, directlyDecreases, and regulates. Correlative relationships include positiveCorrelation and negativeCorrelation.

Correlative

If the observation comes from the comparison of human tumors grouped by the occurrence of a specific mutation, then the relationship should generally be expressed as correlative. In this case there is no experimental perturbation. In this example, most patient tumor samples with an EGFR L858R mutation were observed to exhibit a reduction in ERBB2 tyrosine 1248 phosphorylation compared to wild-type samples (http://www.ncbi.nlm.nih.gov/pubmed/?term=18687633[PMID 18687633]):

p(HGNC:EGFR, var("p.Leu858Arg")) negativeCorrelation p(HGNC:ERBB2, pmod(Ph,Tyr,1248))

In this case, no evidence is presented to suggest that the differences in ERBB2 phosphorylation are causally related to the EGFR mutation, only that the two observations are inversely correlated. Note that the subject and object terms are interchangeable for correlative relationships.

Causal

If the observation comes from the comparison of experimentally controlled states, like gene deletion, overexpression, or introduction of a mutant allele into a cell line or animal, the experimental perturbation can generally be represented as the subject term of a causal statement. In this example, DUSP6 RNA is observed to be upregulated in immortalized human bronchial epithelial cells transfected with EGFR mutant L858R, as compared to WT EGFR (http://www.ncbi.nlm.nih.gov/pubmed/16489012[PMID 16489012]):

p(HGNC:EGFR, var("p.Leu858Arg")) -> r(HGNC:DUSP6)

In this case, the EGFR mutation is introduced as an experimentally-controlled perturbation.

Object Terms (Measurements)

How should I represent microarray data?

We record the results of experiments like microarrays and RT-PCR, which measure RNA abundances, by representing the object terms as RNA abundances. Only significant effects (e.g., meeting minimum criteria for fold change and statistical significance) should be recorded in BEL Statements.

In a causal BEL Statement, the subject term generally represents an experimentally manipulated entity while the object term represents a measured entity. Our general practice is to represent the object terms in BEL Statements with the terms most closely related to the experimental measurement.

This direct representation of the measurement in BEL supports the creation of KAMs to which 'omic data can be mapped directly and analyzed using automated reasoning applications like Whistle. Inference of the potential downstream consequences of RNA expression changes is supported by connection of RNA abundances to the corresponding proteins during knowledge network compilation

Statement Annotations

How do I annotate a relationship observed in multiple biological contexts?

Often, the scientific literature reports a relationship as occurring across several biological contexts.

Our general practice is to represent each observation with a separate statement. Several annotations can be used to describe the same context, e.g., 'lung' and 'fibroblast', but distinct BEL statements should be used to describe each experimental context that the relationship is observed in.

Example

http://www.ncbi.nlm.nih.gov/pubmed/18650932[PMID 18650932] - siRNA knockdown of the atypical PKC-interacting protein Par-4 (PAWR) increases phosphorlyation of AKT at Serine 473 in both human 293 and A549 cells.

"To test whether this is also true in human cells, we used a Par-4 siRNA to deplete endogenous Par-4 levels in human 293 cells and in the A549 human lung adenocarcinoma cell line. Cells were treated with control or Par-4-specific siRNAs, after which they were kept for 24 h in serum-free medium conditions and then stimulated with serum. Data in http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2519103/figure/f5/[Figure 5E and F] clearly demonstrate that the knockdown of Par-4 provokes enhanced serum-activated phospho-Akt-Ser473 levels in A549 and 293 human cells, respectively."

Not recommended:

SET CellLine = {A549, 293}
p(HGNC:PAWR) -| p(SFAM:"AKT Family", pmod(Ph, Ser, 473))

Recommended:

SET CellLine = A549

p(HGNC:PAWR) -| p(SFAM:"AKT Family", pmod(Ph, Ser, 473))

SET CellLine

p(HGNC:PAWR) -| p(SFAM:"AKT Family", pmod(Ph, Ser, 473))

Modified Proteins

  • How do I represent a protein modification when specific information is not available?
  • How do I represent a protein modification within a complex?
  • Xmultphosphforprotein, How do I represent a situation where multiple phosphorylations are required for a protein's activity?
  • How do I represent a situation where one protein modification initiates additional modifications?
  • Xremovalofproteinmod, How do I represent removal of a protein modification (e.g., dephosphorylation, deubiquitination)?

How do I represent a protein modification when specific information is not available?

BEL terms for Protein Modifications, post-translational modifications of proteins specify the type of modification, the modified amino acid, and the position of the modified amino acid. The modified amino acid and position are not required, so protein modifications can be represented with less specific information.

Example

Human AKT1 protein modified by phosphorylation at serine 473

p(HGNC:AKT1, pmod(Ph, Ser, 473))

Human AKT1 protein modified by phosphorylation at an unspecified serine residue

p(HGNC:AKT1, pmod(Ph, Ser))

Human AKT1 protein that has been modified by phosphorylation at an unspecified amino acid residue

p(HGNC:AKT1, pmod(Ph))

As a general rule, if specific information is available, it should be used. In some cases, this involves investigation sections of a paper outside of the evidence text or other referenced papers to determine which specific modifications have been measured.

Non-specific protein modification terms have limited value in the context of a knowledge network. For example, phosphorylation at different sites of the same protein can have opposing effects. For example: "Akt-phosphorylated FOXO interacts with the ubiquitin ligase Skp2 and is targeted for proteasomal degradation" (http://www.ncbi.nlm.nih.gov/pubmed?term=15917664[PMID 15917664])

Example

Recommended:

act(p(SFAM:"AKT Family"), ma(kin)) => (act(p(HGNC:SKP2)) => deg(p(SFAM:"FOXO Family")))

Not recommended:

p(SFAM:"FOXO Family", pmod(Ph)) => (act(p(HGNC:SKP2)) => deg(p(SFAM:"FOXO Family")))

The first BEL Statement indicates that the kinase activity of AKT increases the degradation of FOXO by SKP2. The second statement indicates that phosphorylation of FOXO increases the degradation of FOXO by SKP2. In this case, more information is captured by using the phosphorylating kinase AKT as the subject term instead of the non-specified phosphorylation of FOXO.

How do I represent a protein modification within a complex?

In many cases, complexes include proteins with post-translational modifications and these modifications influence complex formation. For example, HIF1A that has been hydroxylated on proline residues 402 and 564 interacts with VHL (http://www.ncbi.nlm.nih.gov/pubmed?term=17925579[PMID 17925579]).

Our general practice is to represent this type of event as a causal statement in BEL, with the modified protein as the subject term and the complex with no specified modifications as the object term. Because the modified protein is a component of the complex, we use a direct causal relationship:

p(HGNC:HIF1A, pmod(Hy, Pro, 402)) => complex(p(HGNC:HIF1A), p(HGNC:VHL))

p(HGNC:HIF1A, pmod(Hy, Pro, 564)) => complex(p(HGNC:HIF1A), p(HGNC:VHL))

While BEL allows representation of complexes with the modified proteins as components, we do not recommend this approach:

complex(p(HGNC:HIF1A, pmod(Hy, Pro, 402)), p(HGNC:VHL))

The practice of composing a complex using protein abundances without any specified modifications provides a standardized representation for complexes and allows the effects of modifications on complex formation to be captured as causal relationships. The modified forms of HIF1A, p(HGNC:HIF1A, pmod(Hy, Pro, 402)) and p(HGNC:HIF1A, pmod(Hy, Pro, 564)) are considered a subset of the total p(HGNC:HIF1A).

This approach enables representation of the effects of multiple protein modifications on complex formation by using a causal statement for each modification.

BELv2.0 does not provide a specific representation of unmodified protein abundances.

Exception – modified histones bound to the promoter of a specific gene

One exception to our general practice of not specifying protein modifications within complex abundances is the interaction of specific modified histones with the promoter of a specific gene. In this example, cigarette smoke is observed to increase H3K27me3 levels at the DKK1 promoter (http://www.ncbi.nlm.nih.gov/pubmed?term=19351856[PMID 19351856]).

 a(SCHEM:"smoke condensate, cigarette (gas phase)") -> \
    complex(p(PFH:"Histone H3 Family",pmod(Me3, Lys, 27)), g(HGNC:DKK1))

In this example, the modification of the histone by trimethylation does not affect its binding to the gene DKK1. In addition, cigarette smoke does not increase or decrease the overall abundance of the modified histone, only the abundance of the modified histone at the DKK1 promoter.

How do I represent a situation where multiple phosphorylations are required for a protein's activity?

In many cases two or more distinct modifications are required simultaneously for protein activity, and neither modification alone is sufficient. For example, MAPK3 must be phosphorylated at two sites, Threonine 202 and Tyrosine 204, to be active. Our general practice is to take the simple, most general approach, and model the effect of each site separately:

p(HGNC:MAPK3, pmod(Ph, Thr, 202)) => act(p(HGNC:MAPK3), ma(kin))

p(HGNC:MAPK3, pmod(Ph, Tyr, 204)) => act(p(HGNC:MAPK3), ma(kin))

A multiply-modified abundance term can be used if it is of high importance to capture the requirement for both modifications:

p(HGNC:MAPK3, pmod(Ph, Thr, 202)), pmod(Ph, Tyr, 204)) => act(p(HGNC:MAPK3), ma(kin))

How do I represent a situation where one protein modification initiates additional modifications?

In many cases a specific protein modification may be dependent on another modification of the same protein. In this case, the first protein modification can be modeled as the upstream cause of the second. In this example, phosphorylation of CTNNB1 at Serine 45 initiates phosphorylation of CTNNB1 at other sites including Threonine 41 by GSK3 (http://www.ncbi.nlm.nih.gov/pubmed/?term=16618120[PMID 16618120]):

p(HGNC:CTNNB1, pmod(Ph, Ser, 45)) => p(HGNC:CTNNB1, pmod(Ph, Thr, 41))

We represent this relationship as direct, because the subject and object terms have the same root abundance node.

Because the kinase mediating the second phosphorylation is known, this relationship can be modeled alternatively as a nested statement:

p(HGNC:CTNNB1, pmod(Ph, Ser, 45)) => \
    (kin(p(SFAM:"GSK3 Family")) => p(HGNC:CTNNB1, pmod(Ph, Thr, 41)))

How do I represent removal of a protein modification (e.g., dephosphorylation, deubiquitination)?

Removal of a specific protein modification is represented simply as a decrease in the abundance of the modified protein.

Deubiquitination

In this example, STAMBP deubiquitinates F2RL1 protein (http://www.ncbi.nlm.nih.gov/pubmed?term=19684015[PMID 19684015]):

act(p(HGNC:STAMBP)) =| p(HGNC:F2RL1, pmod(Ub))

Deubiquitination is represented simply as a decrease in the ubiquitinated form of the protein. Because in this example STAMBP is the deubiquitinating enzyme, we used a directlyDecreases relationship.

Dephosphorylation

In this example, the phosphatase CDC25C dephosphorylates CDK1 at tyrosine 15 (http://www.ncbi.nlm.nih.gov/pubmed?term=1384126[PMID 1384126]):

act(p(HGNC:CDC25C), ma(phos)) =| p(HGNC:CDK1, pmod(Ph, Tyr, 15))

Similar to deubiquitination, dephosphorylation is represented simply as a decrease in the modified form of the protein.

Reactions

  • How can I represent a reversible metabolic reaction?
  • When and why should I use a reaction term?

How can I represent a reversible metabolic reaction?

A reversible reaction can be represented by modeling the reaction with the products and reactants interchanged.

For example, HSD11B1 acts primarily to convert cortisone to active cortisol, but in some cell types the reverse reaction is favored (http://www.ncbi.nlm.nih.gov/pubmed?term=12530648[PMID 12530648]):

 act(p(HGNC:HSD11B1), ma(cat)) => \
  rxn(reactants(a(CHEBI:NADPH), a(CHEBI:cortisone)), products(a(CHEBI:"NADP(+)"), a(CHEBI:cortisol)))

 act(p(HGNC:HSD11B1), ma(cat)) => \
  rxn(reactants(a(CHEBI:"NADP(+)"), a(CHEBI:cortisol)), products(a(CHEBI:NADPH), a(CHEBI:cortisone)))

The top statement represents the forward reaction and the bottom statement represents the reverse reaction.

When and why should I use a reaction term?

When_ should I use a reaction term

Reaction terms allow the representation of a transformation of a list of reactants into a list of products. In this example, the superoxide dismutase SOD1 converts superoxide to hydrogen peroxide:

act(p(HGNC:SOD1), ma(cat)) => rxn(reactants(a(CHEBI:superoxide)), products(a(CHEBI:"hydrogen peroxide")))

It is not necessary to include all reactants and products, especially if they are ubiquitous small molecules. In the above example the reactant hydrogen and product oxygen have been omitted from the reaction.

Why_ should I use a reaction term

It is possible to represent the above reaction with separate statements linking the activity of SOD1 to decreased abundances of the reactants and increased abundances of the products:

act(p(HGNC:SOD1), ma(cat)) =| a(CHEBI:superoxide)
act(p(HGNC:SOD1), ma(cat)) => a(CHEBI:"hydrogen peroxide")

While this representation describes the function of the catalytic enzyme SOD1, it does not link the product hydrogen peroxide to the reactant superoxide.

Protein-Protein Interactions

How do I represent a physical interaction between two entities?

You can use a complex abundance to represent binding events between two or more abundance terms. The subject term of your BEL Statement will be the complex. BEL Statements do not require a relationship and object term.

The order in which the members of the complex are listed is not important.

Examples

EPOR physically interacts with CSF2RB, the common beta-receptor (http://www.ncbi.nlm.nih.gov/pubmed?term=15456912[PMID 15456912])

 complex(p(MGI:EPOR), p(MGI:CSF2RB))

KEAP1 binds 15-deoxy-Delta(12,14)-prostaglandin J2 (http://www.ncbi.nlm.nih.gov/pubmed?term=15917255[PMID 15917255])

 complex(p(HGNC:KEAP1), a(CHEBI:"15-deoxy-Delta(12,14)-prostaglandin J2"))

KEAP1, CUL3, and RBX1 copurify and are part of a functional E3 ubiquitin ligase complex (http://www.ncbi.nlm.nih.gov/pubmed?term=15572695[PMID 15572695])

 complex(p(HGNC:KEAP1), p(HGNC:RBX1), p(HGNC:CUL3))

The AP-1 transcription complex binds the CCL23 promoter (http://www.ncbi.nlm.nih.gov/pubmed?term=17368823[PMID 17368823])

 complex(p(GOCC:"AP1 complex"), g(HGNC:CCL23))

Protein Families

When should I use a protein family instead of a specific protein?

Protein families can be used to represent protein abundances in cases where the information presented by the source does not allow identification of the specific protein. For example:

Example 1:

SET Citation = "pubmed:11715018"
SET Evidence = "Akt physically associates with MDM2 and phosphorylates it at Ser166 and Ser186."

act(p(fplx:AKT), ma(kin)) => p(HGNC:MDM2, pmod(Ph,Ser,166))
act(p(fplx:AKT), ma(kin)) => p(HGNC:MDM2, pmod(Ph,Ser,186))

Here, Akt may refer to AKT1, AKT2, and/or AKT3.

Example 2:

SET Citation = "pubmed:17003045"
SET Evidence = "We show that Siah2 is subject to phosphorylation by p38 MAPK ... Phosphopeptide mapping identified T24 and S29 as the primary phospho-acceptor sites."

act(p(PFH:"MAPK p38 Family"), ma(kin)) => p(HGNC:SIAH2, pmod(Ph, Thr, 24))
act(p(PFH:"MAPK p38 Family"), ma(kin)) => p(HGNC:SIAH2, pmod(Ph, Ser, 29))

Here, it is not clear which specific p38 MAPK is responsible (MAPK11, MAPK12, MAPK13, or MAPK14).

Example 3:

"Hip encodes a membrane glycoprotein that binds to all three mammalian Hedgehog proteins." (http://www.ncbi.nlm.nih.gov/pubmed?term=10050855[PMID 10050855])

complex(p(SFAM:"Hedgehog Family"), p(MGI:Hhip))

complex(p(MGI:Ihh), p(MGI:Hhip))

complex(p(MGI:Shh), p(MGI:Hhip))

complex(p(MGI:Dhh), p(MGI:Hhip))

In this case, all three hedgehog family members are reported to bind to the hedgehog interacting protein (Hhip). Statements can be modeled using the family as well as each individual member.

More Cookbook

Causal Statement Examples

Causal statements connect subject and object terms with a causal <<Xincreases, increases>>, <<Xdecreases, decreases>>, or <<Xcnc, causesNoChange>> relationship. Subject terms can be an abundance or process (including activities and transformations) and object terms can be either an abundance, a process, or a second BEL Statement.

  • <>
  • <>
  • <>

Causal increase

Example

These statements use the causal <<Xincreases, increases>> relationship. These statements are annotated with a citation and supporting evidence text, as well as with the cell line and species context for the experimental observations represented by the statements. These two statements represent the observation that increases in IL6 protein abundance cause increases in the RNA abundance of ENO1 and XBP1. These statements are annotated with CellLine and Species to indicate that the experimental observation was made in the context of the cell line "U266" and species "9606" (Homo sapiens).

Long Form

SET Citation = {"PubMed", "Int J Oncol 1999 Jul 15(1) 173-8", "10375612"}

SET Support = "Northern blot analysis documented that two transcription factor genes chosen for further study, c-myc promoter-binding protein (MBP-1) and X-box binding protein 1 (XBP-1), were up-regulated in U266 cells about 3-fold relative to the cell cycle-dependent beta-actin gene 12 h after IL-6 treatment"

SET CellLine = "U266"

SET Species = "9606"

// disambiguation MBP-1 = HNGC ENO1
proteinAbundance(HGNC:IL6) increases rnaAbundance(HGNC:ENO1)
proteinAbundance(HGNC:IL6) increases rnaAbundance(HGNC:XBP1)

Short Form

p(HGNC:IL6) -> r(HGNC:ENO1)
p(HGNC:IL6) -> r(HGNC:XBP1)

Causal decrease

Example

This statement demonstrates a causal statement using the <<Xdecreases, decreases>> relationship. The statement expresses that increases in the abundance of corticosteroid molecules cause decreases in the frequency or intensity of the biological process inflammation. This statement is annotated with an Anatomy and Disease to indicate that the relationship was observed in the context of the cardiovascular system and the disease Stroke.

Long Form

 SET Citation = {"PubMed", "J Mol Med. 2003 Mar;81(3):168-74. Epub 2003 Mar 14.", "12682725"}

 SET Support ="high-dose steroid treatment decreases vascular
  inflammation and ischemic tissue damage after myocardial
  infarction and stroke through direct vascular effects involving
  the nontranscriptional activation of eNOS"

 SET Anatomy = "cardiovascular system"

 SET MeSHDisease = "Stroke"

 abundance(CHEBI:corticosteroid) decreases biologicalProcess(MESHD:Inflammation)

Short Form

a(CHEBI:corticosteroid) -| path(MESHD:Inflammation)

Causes no change

The <<Xcnc, causesNoChange>> relationship can be used to record the lack of an observed effect.

Example

The epidermal growth factor receptor (EGFR) ligand amphiregulin (AREG) is observed to increase NF-kappaB transcriptional activity while the EGFR ligand EGF has no effect.These statements express that an increase of AREG protein abundance causes an observed increase in the transcriptional activity of the NF-kappaB complex, and that an increase EGF does not.

Long Form

 SET Citation = {"PubMed", "Mol Cancer Res 2007 Aug 5(8) 847-61", "17670913"}

 SET Support ="Furthermore, EGFR, activated by amphiregulin but not
  epidermal growth factor, results in the prompt activation of the
  transcription factor nuclear factor-kappaB (NF-kappaB)"

 # disambiguation Amphiregulin = HGNC AREG

 proteinAbundance(HGNC:AREG) increases activity(complexAbundance(GOCC:"NF-kappaB complex"), molecularActivity(tscript))

 proteinAbundance(HGNC:EGF) causesNoChange activity(complexAbundance(GOCC:"NF-kappaB complex"), molecularActivity(tscript))

Short Form

p(HGNC:AREG) -> act(complex(GOCC:"NF-kappaB complex"), ma(tscript))
p(HGNC:EGF) causesNoChange act(complex(GOCC:"NF-kappaB complex"), ma(tscript))

Correlative Statement Examples

<> link abundances and biological processes when no causal relationship is known.

  • <>
  • <>

negativeCorrelation

This statement expresses that an increase in cytoplasmic FGF2 protein positively correlates with an increase in the pathology Chronic Obstructive Pulmonary Disease. The subject and object terms of correlative statements are interchangeable. The <<XnegCor, negativeCorrelation>> relationship is used to represent inverse correlative relationships, i.e., a decrease in A is correlated with an increase in B.

 SET Citation = {"PubMed", "J Pathol. 2005 May;206(1):28-38.", "15772985"}

 SET Support ="Quantitative digital image analysis revealed
 increased cytoplasmic expression of FGF-2 in bronchial epithelium
 (0.35 +/- 0.03 vs 0.20 +/- 0.04, p < 0.008) and nuclear
 localization in ASM (p < 0.0001) in COPD patients compared with
 controls."

 SET Tissue = "epithelium"

 proteinAbundance(HGNC:FGF2, location(GOCC:cytoplasm)) positiveCorrelation \
  pathology(MESHD:"Pulmonary Disease, Chronic Obstructive")

association

The direction of causal effect or correlation of two abundance or biological process terms is not always specified. The <<Xassociation, association>> relationship can be used in these cases.

This statement represents that abundance of protein designated by the name Nr2f2 in the MGI namespace is associated in an unspecified manner with the biological process angiogenesis.

Long Form

 SET Citation = {"PubMed", "Mech Ageing Dev. 2004 Oct-Nov;125(10-11):719-32.", "15541767"}

 SET Support ="COUP-TFII is involved in the angiogenic process in the developing embryos."

 # disambiguation - COUP-TFII refers to MGI Nr2f2

 SET MeSHAnatomy = "Embryo, Mammalian"

 proteinAbundance(MGI:Nr2f2) association biologicalProcess(GO:angiogenesis)

Short Form

p(MGI:NR2F2) -- bp(GO:angiogenesis)

Direct Causal Statement Examples

The following examples demonstrate the use of <> in causal statements. The direct causal relationships <<XdIncreases, directlyIncreases>> and <<XdDecreases, directlyDecreases>> are special forms of the causal <<Xincreases, increases>> and <<Xdecreases, decreases>> relationships where the mechanism of the causal relationship involves the physical interaction of entities related to the BEL Statement subject and object terms.

  • <>
  • <>
  • <>
  • <>
  • <>

Example - Ligand and Receptor

In this example, the <<XdIncreases, directlyIncreases>> relationship is used to represent activation of a receptor by its ligand. This statement expresses that amphiregulin (AREG) activates its receptor, the Epidermal Growth Factor Receptor (EGFR). This relationship is direct because ligands directly interact with their receptors.

Long Form

SET Citation = {"PubMed", "Mol Cancer Res 2007 Aug 5(8) 847-61", "17670913"}
SET Support ="Furthermore, EGFR, activated by amphiregulin"
// disambiguation Amphiregulin = HGNC AREG
// EGFR is known to have kinase activity
proteinAbundance(HGNC:AREG) directlyIncreases activity(proteinAbundance(HGNC:EGFR), molecularActivity(kin))

Short Form

p(HGNC:AREG) => act(p(HGNC:EGFR), ma(kin))

Example - Kinase and Substrate

In this example, the <<XdIncreases, directlyIncreases>> relationship is used to represent the phosphorylation of a protein substrate by a kinase. This statement expresses that the kinase activity of CDK1 protein causes an increase in the modification of FOXO1 protein by phosphorylation at serine 249. The relationship is direct because the kinase physically interacts with its target.

Long Form

SET Citation = {"PubMed", "Science 2008 Mar 21 319(5870) 1665-8.", "18356527"}
SET Support ="We found that Cdk1 phosphorylated the transcription factor FOXO1 at Ser249 in vitro and in vivo."
activity(p(HGNC:CDK1), ms(kin)) => p(HGNC:FOXO1, pmof(Ph, Ser, 249))

Example - Catalyst and Reaction

In this example, the direct activation of a reaction by a catalytic enzyme is represented. The statement indicates that an increase in the catalytic activity of ALOX5 increase the transformation of the reactant '5(S)-HPETE' to the products 'leukotriene A4' and 'water'. The relationship is considered direct because ALOX5 protein is the catalyzing enzyme.

Long Form

 SET Citation = {"Other", "Reactome: Leukotriene synthesis", "REACT_15354.1"}

 SET Support ="Dehydration of 5-HpETE to leukotriene A4. In the
  second step, 5-lipoxygenase converts 5-HpETE to an allylic
  epoxide, leukotriene A4."

 activity(proteinAbundance(HGNC:ALOX5), molecularActivity(cat)) directlyIncreases \
  reaction(reactants(abundance(CHEBI:"5(S)-HPETE")), \
  products(abundance(CHEBI:"leukotriene A4"), abundance(CHEBI:water)))

Short Form

 act(p(HGNC:ALOX5), ma(cat)) => rxn(reactants(a(CHEBI:"5(S)-HPETE")), products(a(CHEBI:"leukotriene A4"), a(CHEBI:water)))

Example - Self-Referential Relationships

In this example, the <<XdDecreases, directlyDecreases>> relationship is used to represent the effect of a protein modification on the activity of the same protein. These statements express that the modification of GSK3A and GSK3B protein by phosphorylation on serines 9 and 21, respectively, inhibits the activity of GSK3A and GSK3B. These relationships are considered direct, because they are self-referential. The modification of the protein abundance by phosphorylation inhibits the activity of the same protein abundance.

Long Form

 SET Citation = {"PubMed", "Proc Natl Acad Sci U S A 2000 Oct 24 97(22) 11960-5", "11035810"}

 SET Support ="GSK-3 activity is inhibited through phosphorylation
  of serine 21 in GSK-3 alpha and serine 9 in GSK-3 beta."

 proteinAbundance(HGNC:GSK3A, proteinModification(Ph, Ser, 21)) \
  directlyDecreases activity(proteinAbundance(HGNC:GSK3A))

 proteinAbundance(HGNC:GSK3B, proteinModification(Ph, Ser, 9)) \
  directlyDecreases activity(proteinAbundance(HGNC:GSK3B))

Short Form

p(HGNC:GSK3A, pmod(Ph, S, 21)) =| act(p(HGNC:GSK3A))
p(HGNC:GSK3B, pmod(Ph, S, 9)) =| act(p(HGNC:GSK3B))

Example - Direct Transcriptional Control

In this example, the direct activation of a RNA transcription is encoded. The statement expresses that increases in the transcriptional activity of FOXO1 protein directly increase the RNA abundance of CEBPB. This relationship is considered direct because the transcription factor, FOXO1, directly binds the promoter of the CEBPB gene, increasing the expression of CEBPB RNA.

Long Form

SET Citation = {"PubMed", "Biochem Biophys Res Commun. 2009 Jan 9;378(2):290-5. Epub 2008 Nov 21.", "19026986"}

SET Support ="We found that Foxo1 increased the expression of CCAAT/enhancer binding protein (C/EBPbeta, a positive regulator of monocyte chemoattractant protein (MCP)-1 and interleukin (IL)-6 genes, through directly binding to its promoter."

activity(proteinAbundance(HGNC:FOXO1), molecularActivity(tscript)) directlyIncreases rnaAbundance(HGNC:CEBPB)

Short Form

act(p(HGNC:FOXO1), ma(tscript)) => r(HGNC:CEBPB)

Nested Statement Example

This example demonstrates use of a nested causal statement in which the object of a causal statement is itself a causal statement. In the relationship described by the evidence text, CLSPN specifically increases the activity of ATR to phosphorylate the target protein CHEK1 and does not affect the kinase activity of ATR towards its other targets. The use of the nested statement allows the representation of the information that CLSPN increases the phosphorylation of CHEK1 via the kinase activity of ATR, without incorrectly indicating that CLSPN generally increases the kinase activity of ATR.

Long Form

SET Citation = {"PubMed", "Mol Cell Biol 2006 Aug 26(16) 6056-64.", "16880517"}

SET Species = "9606"

SET Support ="Consistently, the RNAi-mediated ablation of Claspin selectively abrogated ATR's ability to phosphorylate Chk1 but not other ATR targets."

proteinAbundance(HGNC:CLSPN) increases (activity(proteinAbundance(HGNC:ATR), molecularActivity(kin)) directlyIncreases proteinAbundance(HGNC:CHEK1, proteinModification(Ph)))

Short Form

p(HGNC:CLSPN) -> (act(p(HGNC:ATR), ma(kin)) => p(HGNC:CHEK1, pmod(Ph)))

Other Examples

Membership Assignment Examples

These examples demonstrate the assignment of members to groups. Because all BEL terms denote classes, membership in a group is an important special case where subsets of a class that define the class are designated.

[TIP]

The BEL Framework adds family members to protein families and complex components to named complexes during network compilation.

  • <>
  • <>

Protein Family

In this example, members of a protein family are assigned using the <<hasMember>> and <<hasMembers>> relationships.

The hasMembers relationship is used to assign a list of protein abundances as members of a protein family. This relationship is a syntactic convenience that is equivalent to the set of two statements using the hasMember relationship. These statements designate the protein abundances of MAPK8 and MAPK9 as members of the JNK MAPK protein family. The term representing the JNK family is a protein abundance based on the name MAPK JNK Family in the Selventa Protein Families namespace. p(SFAM:"MAPK JNK Family") hasMembers list(p(HGNC:MAPK8), p(HGNC:MAPK9))

The hasMember relationship is used to assign individual protein abundances to a protein family. p(SFAM:"MAPK JNK Family") hasMember p(HGNC:MAPK8) p(SFAM:"MAPK JNK Family") hasMember p(HGNC:MAPK9)

Complex Component

In this example components are assigned to a named protein complex using the <> and <> relationships.

The hasComponents relationship is similar to the hasMembers relationship and is used to assign a list of abundances as components of a complex.These statements designate the protein abundances of RAD9A, RAD1, and HUS1 as components of the complex abundance of the checkpoint clamp complex.

complex(GOCC:"checkpoint clamp complex") hasComponents list(p(HGNC:RAD9A), p(HGNC:RAD1), p(HGNC:HUS1))

The hasComponent relationship is used to assign individual abundances to a named protein complex.

complex(GOCC:"checkpoint clamp complex") hasComponent p(HGNC:RAD9A)

complex(GOCC:"checkpoint clamp complex") hasComponent p(HGNC:RAD1)

complex(GOCC:"checkpoint clamp complex") hasComponent p(HGNC:HUS1)

The single hasComponents statement is equivalent to the set of three hasComponent statements.

Example - Binding Interaction

The complexAbundance() function can be used to specify molecular interactions between abundances. This function can take either a list of abundances that define a molecular complex or a namespace value that represents a molecular complex (e.g., many GO Cellular Component values) as an argument. These examples demonstrate the use of the complexAbundance() function to represent protein-protein, protein-chemical, and protein-DNA interactions.

  • Protein – protein interactions
  • Protein – DNA interactions
  • Protein – small molecule interactions

Example - protein-protein interaction as BEL statement

This statement represents that MTOR and AKT1S1 proteins physically interact. Note that this statement has only an object term and no subject term and relationship.

SET Citation = {"PubMed", "17277771"}
SET Support = "Here, we identify PRAS40 (proline-rich Akt/PKB substrate 40 kDa) as a novel mTOR binding partner"
// disambiguation PRAS40 = HGNC AKT1S1
complex(p(hgnc:391 ! AKT1S1), p(hgnc:3942 ! MTOR))

Example - protein-protein interaction as Statement object

Here, a protein-protein interaction is the object of a BEL Statement. This statement expresses that the MTOR and STAT3 proteins associate and that increases in the protein abundance of BMP4 can increase the abundance of the complex comprised of MTOR and STAT3.

SET Citation = {"PubMed", "12796477"}
SET Support = "Upon BMP4 treatment, the serine-threonine kinase FKBP12/rapamycin-associated protein (FRAP), mammalian target of rapamycin (mTOR), associates with Stat3 and facilitates STAT activation."
p(hgnc:1071 ! BMP4) -> complex(p(hgnc:3942 ! MTOR), p(hgnc:11364 ! STAT3))

Protein – DNA interactions

Example - transcription factor protein binding to DNA

This statement expresses that STAT3 protein binds to the CCL11 gene DNA, and that this association is increased by IL17A.

SET Citation = {"PubMed", "19265112"}
SET Support ="IL-17A induced at 1 h a marked enrichment of STAT3- associated CCL11 promoter DNA"
p(hgnc:5981 ! IL17A) -> complex(p(hgnc:11364 ! STAT3), g(hgnc:10610 ! CCL11))

Protein – small molecule interactions

Example - protein binding to a small molecule

This statement represents that PIP3 binds AKT1 protein.

SET Citation = {"PubMed", "15987444"}
SET Support = "After PIP3 binding, Akt1 is activated"
// disambiguation PIP3 = CHEBI 1-phosphatidyl-1D-myo-inositol 3,4,5-trisphosphate
complex(a(CHEBI:"1-phosphatidyl-1D-myo-inositol 3,4,5-trisphosphate"), p(hgnc:391 ! AKT1))

Causal

p(HGNC:CCND1) => act(p(HGNC:CDK4))

The abundance of the protein designated by CCND1 in the HGNC namespace directly increases the activity of the abundance of the protein designated by CDK4 in the HGNC namespace.

Causal

p(HGNC:BCL2)-| bp(MESHPP:Apoptosis)

The abundance of the protein designated by BCL2 in the HGNC namespace decreases the biological process designated by apoptosis in the MESHPP (phenomena and processes) namespace.