Genetic Interactions/OldNotes

This page contains notes and implementation details of features that were present in the original genetic interaction proposal. Many of these features were considered too detailed for a first version and were not included. These notes will be useful when extending the representation of genetic interactions.

Chemical Genetic Interactions
Requirements:
 * 1) Small molecule-gene genetic interactions.
 * 2) Small molecule-small molecule genetic interactions.

Studying genetic interactions involving small molecules is often called chemical genetics.

Open issues:
 * 1) For chemical genetic interactions, (how) should we include concentration response profiles?

Notes:
 * 1) Can we better define the relationship between genes and small molecules? - Anchor(smallMoleculeNote)
 * 2) * Small molecules can target genes. The possibilities are a) the target is unknown, b) there is more than one target in the system or c) there is a single specific target in the system. For case a), we can only use the small molecule system perturbation as a participant in the genetic interaction. For case b), we can't link the small molecule to the target genes for the purposes of defining a genetic interaction, since there is no way to define a genetic interaction when you don't know which of the target genes are involved (although you could limit the number of candidate genes that could be involved and say 'genetic interaction of A with one or more of 'B, C or D'). Thus, for this case, it is likely safe to still use the small molecule system perturbation as an interaction participant. For case c), we know the specific gene involved, thus it is the preferred participant in the interaction. In this case, the small molecule is part of the 'experimental form' of the gene, analogous to a mutation or knock-out.
 * 3) Small molecules can perturb the system by e.g. binding to the gene product in a way that affects the gene function (e.g. by blocking a natural binding surface or active site). Small molecules have issues of specificity i.e. they may perturb multiple genes at the same time. However, they can be highly specific, targeting a single function of a multi-function gene and they may be more easily (than a mutation) and reversibly introduced into a biological system. Small molecules may also have indirect effects, where they are metabolized in by the biological system to produce the actual active chemical.

Background reading:
 * Chemical genomic profiling of biological networks using graph theory and combinations of small molecule perturbations - J Am Chem Soc. 2003, Haggarty SJ, Clemons PA, Schreiber SL
 * Integration of chemical-genetic and genetic interaction data links bioactive compounds to cellular target pathways - Nature Biotechnology 2004, Parsons AB, Brost RL, Ding H, Li Z, Zhang C, Sheikh B, Brown GW, Kane PM, Hughes TR, Boone C

Present in original implementation
Genetic interaction types:
 * Intergenic non-complementation interaction: two perturbations fail to complement but act on different genes. This could indicate that both genes are physically interacting. See.
 * More specialized genetic interaction terms already available in the PSI-MI controlled vocabulary for genetic interaction types could be used. The important ones for current pathway databases are synthetic-lethal and synthetic-sick, since these are the largest interaction class. See.

Evidence:
 * Property: PHENOTYPE-INEQUALITY, range = phenotypeInequality (optional), property definition: "The inequality relation of phenotypes of all perturbations in the interaction. More information can be found in: 'Derivation of genetic interaction networks from quantitative phenotype data' by Drees BL et al. Genome Biology 2005."
 * Restriction: Maximum cardinality of PHENOTYPE-INEQUALITY = 1
 * Add a new subclass of the existing utilityClass, called "phenotypeInequality", defined as follows:
 * Class definition: "The inequality relation of phenotypes of all genetic perturbations in the interaction. More information can be found in: 'Derivation of genetic interaction networks from quantitative phenotype data' by Drees BL et al. Genome Biology 2005."
 * Property: INEQUALITY, range = String, property definition: "The inequality relation of phenotypes of all perturbations in the interaction. More information can be found in: 'Derivation of genetic interaction networks from quantitative phenotype data' by Drees BL et al. Genome Biology 2005. E.g. AB<WT=A=B. The inequality must be written as a string using only the less than (<) character, the equal (=) character, the 'WT' symbol and symbols representing the phenotypes caused by individual perturbations represented by the referring genetic interaction instance participants (separated by spaces). A gene participant represents the gene that is perturbed and a small molecule participant represents a small molecule perturbant. Thus, the participants in this property must occur in the referring genetic interaction instance. The symbols used in this string must be defined in the PHENOTYPE-SYMBOL property. Defined symbols (i.e. other than 'WT') can be directly concatenated to represent multiple perturbations occurring simultaneously."
 * Restriction: Cardinality of INEQUALITY = 1
 * Property: PHENOTYPE-SYMBOL, range = phenotypeSymbol, property definition: "The definition of the user-defined symbols used in the phenotype inequality."
 * Restriction: Minimum cardinality of PHENOTYPE-SYMBOL = 2
 * Add a new subclass of the existing utilityClass, called "phenotypeSymbol", defined as follows:
 * Class definition: "The definition of the symbols used in the phenotype inequality defined in the referring phenotypeInequality instance. More information can be found in: 'Derivation of genetic interaction networks from quantitative phenotype data' by Drees BL et al. Genome Biology 2005. Instances of this class may be shared between multiple phenotypeInequality instances to avoid having to redefine many identical phenotypeSymbol instances."
 * Property: GENETIC-PARTICIPANT, range = {gene or smallMolecule}, property definition: "The genetic perturbation represented as a symbol in the referring phenotypeInequality instance PHENOTYPE-INEQUALITY property. The genetic perturbation is represented by the referring genetic interaction instance participants. A gene participant represents the gene that is perturbed and a small molecule participant represents a small molecule perturbant. Thus, the participants in this property must occur in the referring genetic interaction instance."
 * Restriction: Cardinality of GENETIC-PARTICIPANT = 1
 * Property: SYMBOL, range = String, property definition: "A string expressing the symbol e.g. 'A', 'B' (not including the single quote mark). The string must not include spaces or any characters from the set '=<' or the symbol 'WT' (not including the single quote marks)."
 * Restriction: Cardinality of SYMBOL = 1

Notes:
 * 1) Phenotypes can be qualitative (e.g. lethal/viable) or quantitative (e.g. growth rate, amount of red coloring).  This proposal follows the Drees et al. phenotype comparison scheme (e.g. A = B = WT < AB) that is general for both qualitative and quantitative phenotypes.  Further details about quantitative phenotypes are not covered by this proposal. - Anchor(dreesNote)
 * 2) The Drees et al. phenotype comparison scheme (e.g. A = B = WT < AB) only describes interactions between 2 genes, however the scheme is general to any genetic perturbation and is general to interactions involving more than 2 genetic perturbations.  For instance, synthetic lethality between 3 genes could be represented as WT = A = B = C = AB = BC = AC < ABC.
 * 3) Note that the inequality can be dependent on the phenotype description e.g. 'rate of growth' versus 'rate of death'
 * 4) Use of some genetic interaction types may lead to duplicated information in the geneticEvidence class. This could cause errors in data exchange if the information stored in each location disagrees. - Anchor(duplicationNote)
 * 5) * The phenotypeInequality class may duplicate information for some genetic interaction types, like synthetic, because these may only map to single phenotype inequalities.

Closed issues:
 * 1) How much information about the experimental definition of phenotype and the mathematical model of genetic interaction should be encoded? We will only encode minimal information until these concepts are better defined and used more often in the community.
 * 2) Is the genetic interaction gene the same as a gene regulation gene?  Yes.
 * 3) How should the phenotype description be captured if we don't have a controlled vocabulary available? Users should be allowed to create their own controlled vocabulary as long as they follow community accepted best practices, such as use of OBO.
 * 4) What is a practical definition of gene for BioPAX? (also an open issue in the gene regulation proposal). The definition of a gene will not be tackled by BioPAX and will be left to the Sequence Ontology group.
 * 5) What is the organism of a viral gene in a host? It is the virus, not the host.
 * 6) Should more detail on genetic background be stored? E.g. the backgrounds of each parent strain in addition to the child (result) strain? Also, a definition in terms of existing gene instances e.g. wt202 strain -his3 gene +ura3 gene.  No - we will store minimal information about genetic background until more standard representation schemes are available.
 * 7) (How) should quantitative phenotype information be stored? It will not be stored in Level 3
 * 8) Do we need to create specialized genetic interaction types that include phenotype? E.g. synthetic lethal - No - phenotype will not be included in the interaction type due to possible combinatorially large number of interaction types being created.

Alternative representations

 * 1) An alternative way to store mutations is to use state variables. This would have the following effects:
 * 2) The meaning of the participants in the genetic interaction would change to "the participants represent perturbations to a biological system". This is simpler than the current definition.
 * 3) The EXPERIMENTAL-FORM would no longer be used, leading to a change of the geneticEvidence class (add a restriction not to use this property)
 * 4) A GENETIC-EXPERIMENTAL-FORM property would be added to the geneticEvidence class to store a symbol list of the genetic mutation terms (e.g. knock-out)
 * 5) The physicalEntityParticipant or physicalEntityInState class would need to be used for participants, leading to more complexity.
 * 6) The rationale for storing gene perturbations in the experimental form property, instead of e.g. as gene 'state variables': the gene perturbation is an experimental detail. The fact that a genetic interaction is observed with a given set of gene perturbations is enough to define the interaction. Other perturbations are possible that lead to the same experimental conclusion.
 * 7) An alternative way to store the phenotype inequality, which is more complex than the current proposal but potentially easier to compute with, is :

1. Add a new subclass of the existing utilityClass, called "phenotypeInequality", defined as follows: * Class definition: "The inequality relation of phenotypes of all perturbations in the interaction. More information can be found in: "Derivation of genetic interaction networks from quantitative phenotype data" by Drees BL et al. Genome Biology 2005." * Property: EQUAL-PHENOTYPE, range = {gene or smallMolecule}, property definition: "The perturbations that result in equal phenotypes (to all others in the property), where perturbations are represented by the referring genetic interaction instance participants. A gene participant represents the gene that is perturbed and a small molecule participant represents a small molecule perturbation. Thus, the participants in this property must occur in the referring genetic interaction instance." * Property: UNEQUAL-PHENOTYPE, range = unequalPhenotype, property definition: "The perturbations pairs that result in unequal phenotypes (to each other), where perturbations are represented by the referring genetic interaction instance participants. A gene participant represents the gene that is perturbed and a small molecule participant represents a small molecule perturbant. Thus, the participants in this property must occur in the referring genetic interaction instance." * Property: RELATION-TO-WILDTYPE, range = relationToWildType, property definition: "The relationship of the phenotype resulting from one perturbation"

1. Add a new subclass of the existing utilityClass, called "unequalPhenotype", defined as follows: * Class definition: "A single perturbation pair that results in unequal phenotypes (to each other), where perturbations are represented by the referring genetic interaction instance participants. A gene participant represents the gene that is perturbed and a small molecule participant represents a small molecule perturbation. Thus, the participants in the properties of this class must occur in the referring genetic interaction instance." * New property: LESSER-PHENOTYPE, range = {gene or smallMolecule}, property definition: "The lesser of the two phenotypes..." * New property: GREATER-PHENOTYPE, range = {gene or smallMolecule}, property definition: "The greater of the two phenotypes..."

1. Add a new subclass of the existing utilityClass, called "relationToWildType", defined as follows: * Class definition: "The relationship of perturbed phenotypes to the wild-type." * New property: EQUAL-TO, range = {gene or smallMolecule}, property definition: "Wild type phenotype is equal to these participants..." * New property: LESS-THAN, range = {gene or smallMolecule}, property definition: "Wild type phenotype is less than these participants..."