Genetic Interactions

-

= Executive Summary =

Genetic interactions between genes occur when two genetic perturbations (e.g. gene mutations) have a combined phenotypic effect not caused by either perturbation alone. For example, a synthetic lethal interaction occurs when cell growth is possible without either gene A OR Z, but not without both gene A AND Z. If you knock out A and Z together, the cell will die.



Other types of non-lethal genetic interactions are possible. Suppose absence of gene Z shows no phenotype, while absence of A causes a non-lethal fitness defect. Absence of both genes may either mitigate ('buffer') or exacerbate ('aggravate') the phenotype observed when only A is missing. If that is the case, A and Z are said to have a genetic interaction.

A small molecule introduced to the biological system can be a genetic perturbation generally via inactivation of the gene product, though the gene target is generally not known. In this case, we can record the interaction between any known genes and any introduced small molecules.

(See )

- = Introduction =

Goals/Motivation
To represent genetic interactions, as stored in BIND, GRID, MIPS CYGD or Flybase. Genetic interaction data is useful for genetic interaction network analysis (e.g. GenePath), for mapping biological pathways, for prediction of gene function and in the design of single and combination drug therapy for human disease.

= Biological Questions / Use Cases =


 * 1) Genetic interactions resulting from the mutation of two or more independently segregating genes measured by the effect on a single phenotype. e.g. Genetic interactions, including those from SGA high-throughput screens.

= Requirements =

User Requirements
Basic requirements:

We must be able to capture:
 * 1) A list of genetic perturbations (>1) that are involved in affecting a phenotype. Perturbations must specifically affect a known gene e.g. gene mutation or small molecule introduced to the biological system or organism under study specifically known to affect a single gene.
 * 2) The type of the genetic interaction e.g. synthetic, asynthetic, suppressive, epistatic, conditional, additive, single-nonmonotonic, double-nonmonotonic, enhanced, non-complementation.

Experimental evidence requirements:

We must be able to capture:
 * 1) The phenotype that is used to detect that an interaction occurs (can be qualitative or quantitative).
 * 2) The genetic background of the organism the experiment was performed in.
 * 3) The experimental system that was used to determine the genetic interaction.

The above requirements are drawn from the following database representations and data records: (see http://sf.net/projects/biopax/files/other/bind.asn.txt) or Browse
 * BIND - From the BIND-genetic-experiment data type and linked data type of the BIND schema
 * Flybase
 * GRID and MIPS CYGD maintain only gene pairs, type of genetic interaction and a publication.

Practical Software Requirements

 * 1) None specific. Similar to the basic requirements of the rest of BioPAX.

= Proposed Implementation =

(To fulfill the basic requirements)
 * 1) Add a new subclass of interaction called "geneticInteraction", sibling of physicalInteraction, defined as:
 * 2) * Class definition: "Genetic interactions between genes occur when two genetic perturbations (e.g. mutations) have a combined phenotypic effect not caused by either perturbation alone. This is not a physical interaction, but rather logical. For example, a synthetic lethal interaction occurs when cell growth is possible without either gene A OR Z, but not without both gene A AND Z. If you knock out A and Z together, the cell will die. A gene participant in a genetic interaction represents the gene that is perturbed."
 * 3) * Restriction: All PARTICIPANTS instances of class gene
 * 4) * Restriction: All EVIDENCE instances of class geneticEvidence (this requires us to add geneticEvidence to the range of the EVIDENCE property and restrict its use in physicalInteraction to 'evidence' instances)
 * 5) * Restriction: Minimum cardinality of PARTICIPANTS = 2
 * 6) * Move property definition: INTERACTION-TYPE from the physicalInteraction class to the interaction class (move property to the superclass) so that it is inherited by both physical and genetic interaction classes.
 * 7) * The INTERACTION-TYPE property uses the PSI-MI interaction type controlled vocabulary (CV). The PSI-MI interaction type CV needs more genetic interaction types to support this proposal. Ensure that PSI-MI controlled vocabulary contains the following genetic interaction types (requires communication with the PSI-MI group, via their sourceforge feature request tracker).
 * 8) ** The following types and definitions are taken from Drees et al. Genome Biol. 2005;6(4):R38. See . Given phenotypes resulting from genetic perturbations: A = phenotype resulting from genetic perturbation A, and B = phenotype resulting from genetic perturbation B, AB = phenotype resulting from both genetic perturbations together, WT = wild-type phenotype, we have:
 * 9) *** Synthetic interaction: A and B have no effect on the WT background, but the AB combination has an effect.
 * 10) *** Asynthetic interaction: A, B, and the AB combination all have the same effect on the WT background.
 * 11) *** Suppressive interaction: A has an effect on WT, but that effect is abolished by adding the suppressor B, which itself shows no single-mutant effect (for example, WT = B = AB < A); or, the corresponding holds under exchange of A and B.
 * 12) *** Epistatic interaction: A and B have different effects (in terms of direction or magnitude) on the wild-type background and the double mutant has the same phenotype as either A or B (for example, A < WT < B = AB).
 * 13) *** Conditional interaction: A has an effect only in the B background, or the B mutant has an effect only in the A background.
 * 14) *** Additive interaction: Single-mutant effects combine to give a double-mutant effect as per WT < A= B < AB, B < WT = AB < A, WT < A < B < AB, B < WT < AB < A, and all additional inequalities obtained by interchanging A and B, or reversing the effect of both A and B.
 * 15) *** Single-nonmonotonic interaction: B shows opposing effects in the WT and A backgrounds (for example, B > WT and AB< A); or, A shows opposing effects in the WT and B backgrounds, but not both.
 * 16) *** Double-nonmonotonic interaction: Both A and B show opposing effects in the WT background and the background with the other mutant gene.
 * 17) * The following supplemental useful types and definitions are taken from the BIND database specification:
 * 18) ** Enhancement interaction: the A perturbation enhances the phenotype of the B perturbation, or vice versa (e.g. WT = A < B < AB or WT = B < A < AB). This could be conditional or additive by the above scheme.
 * 19) Add a new subclass of physicalEntity called "gene" (sibling to protein, etc.) defined as:
 * 20) * Class definition: "A physical entity that encodes information that can be inherited through replication i.e. DNA or RNA. This is a generalization of the prokaryotic and eukaryotic notion of a gene."
 * 21) * New property: GENE-PRODUCT, range = {protein or RNA}, property definition: "The end-product of a gene, e.g. a protein or microRNA. A gene can have multiple gene products."
 * 22) * Existing property: ORGANISM, range = bioSource
 * 23) * Restriction: Maximum cardinality of ORGANISM = 1
 * 24) * See

(To fulfill the experimental evidence requirements)
 * 1) Add a new subclass of the existing utilityClass class, called "geneticEvidence", defined as:
 * 2) * Class definition: "The support for the existence of a genetic interaction. XREF may reference a publication describing the experimental evidence using a publicationXref or may store a description of the experiment in an experimental description database using a unificationXref (if the referenced experiment is the same) or relationshipXref (if it is not identical, but similar in some way e.g. similar in protocol). Evidence is meant to provide more information than just an xref to the source paper. Examples: A description of a synthetic lethal experiment. The experimental form stores the gene perturbation (e.g. mutation) that resulted in the phenotype. The gene participant must exist in the genetic interaction referencing this geneticEvidence instance."
 * 3) * Property: PHENOTYPE, range = openControlledVocabulary (optional), property definition: "The phenotype measured in the experiment e.g. growth rate, viability."
 * 4) ** Restriction: Maximum cardinality of PHENOTYPE = 1. (See )
 * 5) Property: GENETIC-BACKGROUND, range = geneticBackground (optional), property definition: "The genetic background of the organism used in the experiment."
 * 6) * Restriction: Maximum cardinality of GENETIC-BACKGROUND = 1
 * 7) Add existing properties used in the 'evidence' class: CONFIDENCE, EVIDENCE-CODE, EXPERIMENTAL-FORM, XREF. (See )
 * 8) * Change range of EXPERIMENTAL-FORM property (used by existing evidence class) to {experimentalForm or geneticExperimentalForm}
 * 9) ** Add to existing evidence class in response to this range change - Restriction: All values of EXPERIMENTAL-FORM property are of type experimentalForm.
 * 10) Restriction: All values of EXPERIMENTAL-FORM property are of type geneticExperimentalForm
 * 11) Add a new sublass of utilityClass, called "geneticBackground", defined as:
 * 12) * Class definition: "A genetic background. The genetic mutations from wild-type that are considered background for an experiment. It is usually assumed that background mutations do not affect the results of the experiment, though this is not always the case, thus it is useful to record the background. Defined according to the standards of the field of study. For example, the Saccharomyces cerevisiae community may use notation similar to 'R0013: MAT-alpha hsc82delta::LEU2 hsp82ts-URA3 can1delta::MFA1pr-HIS3 ura3delta0 leu2delta0 his3delta1 lys2delta0 met15delta0' while the Caenorhabditis elegans community may use different nomenclature. The description of the genetic background is stored in the COMMENT property."
 * 13) Add a new subclass of utilityClass, called "geneticExperimentalForm", defined as:
 * 14) * Class definition: "The form of a genetic interaction experiment participant, as it may be modified for purposes of experimental design. For example, a gene mutation or knock-out."
 * 15) * Property: GENETIC-PARTICIPANT, range = gene, property definition: "The gene participant in the referring genetic interaction instance that is perturbed."
 * 16) ** Restriction: Cardinality of GENETIC-PARTICIPANT = 1
 * 17) New property: GENETIC-EXPERIMENTAL-FORM-TYPE, range = Symbol (owl:oneOf), String {"knock-out", "knock-down", "hypermorph", "hypomorph", "antimorph", "over-expressed", "under-expressed"}, property definition: "Descriptor of this experimental form. Terms are:
 * 18) * knock-out: The gene has been completely removed e.g. by genetic engineering
 * 19) * knock-down: The gene expression has been significantly reduced compared to wild-type by introduction of an external substance, e.g. by RNA interference
 * 20) * hypermorph: The gene function has been partially improved compared to wild-type by altering its sequence.
 * 21) * hypomorph: The gene function has been partially reduced compared to wild-type by altering its sequence e.g. a temperature sensitive mutant
 * 22) * antimorph: The gene function has been antagonized by a mutation in another copy of the gene.
 * 23) * amorph: The gene function has been abolished by mutation, though the type of mutation is not known.
 * 24) * mutation: The gene is mutated in some unknown manner.
 * 25) * over-expressed: The gene expression has been significantly increased compared to wild-type by engineering, e.g. by replacing the normal gene promoter with one that overexpresses the gene.
 * 26) * under-expressed: The gene expression has been significantly decreased compared to wild-type by engineering, e.g. by replacing the normal gene promoter with one that underexpresses the gene."
 * 27) ** Restriction: Cardinality of EXPERIMENTAL-FORM-TYPE = 1
 * 28) Add existing property: NAME, range = string, property definition: "The preferred full name for this entity."

= Worked Examples =

Note: the worked examples are being updated so are not totally in sync with the current proposal


 * 1) The proposed OWL implementation (built on BioPAX Level 2 v1.0)
 * 2) * http://sf.net/projects/biopax/files/other/geneticInteractions-biopax.owl
 * 3) * http://sf.net/projects/biopax/files/other/geneticInteractions-biopax.pprj (The Protege project file, which makes the OWL file easier to load into Protege - you need to download both and put them in the same directory to use the .pprj file)
 * 4) A synthetic lethal interaction from BIND - KAR3 genetically interacts with BUD14
 * 5) * http://sf.net/projects/biopax/files/other/geneticInteractions-BIND-eg-biopax.owl
 * 6) * http://sf.net/projects/biopax/files/other/geneticInteractions-BIND-eg-biopax.pprj
 * 7) * More information: Genetic_Interactions/BINDWorkedExample
 * 8) Genetic interaction from BioGRID - ATS1 interactions
 * 9) * http://sf.net/projects/biopax/files/other/geneticInteractions-GRID-eg-biopax.owl
 * 10) * http://sf.net/projects/biopax/files/other/geneticInteractions-GRID-eg-biopax.pprj
 * 11) * More information: Genetic_Interactions/BioGRIDWorkedExample
 * 12) A genetic interaction from Flybase - wg(en11) interactions
 * 13) * http://sf.net/projects/biopax/files/other/geneticInteractions-Flybase-eg-biopax.owl
 * 14) * http://sf.net/projects/biopax/files/other/geneticInteractions-Flybase-eg-biopax.pprj
 * 15) * More information: Genetic_Interactions/FlybaseWorkedExample

How to view worked examples

= Open Issues =


 * 1) Is there a good way to store directionality for a genetic interaction that is compatible with triple mutants?  Genetic interactions involving 2 genes can be directional, such as in epistasis, but this doesn't make sense for triple mutants. It may be possible to introduce a role property in the interaction or experimental description that can tag the participants as source vs. target or upstream vs. downstream.
 * 2) geneticBackground is currently implemented as a class to address earlier discussion comments about extensibility and the issues with representing it as a string.  Should this simply be represented as a string in the geneticEvidence class and we can replace it with an object property later if we need to at a cost of backwards incompatibility at a later date?
 * 3) What other groups are interested in genetic interaction exchange?
 * 4) PSI-MI recently added CV terms to support this proposal. We need to evaluate these terms in relation to this current proposal.
 * 5) After evaluating the recently added PSI-MI terms, the proposal can be simplified, since less new classes are required to capture concepts now encoded in PSI-MI.  Specifically:
 * 6) * the geneticExperimentalForm class can be removed from the proposal
 * 7) * the geneticEvidence class can be removed, since it is redundant with the existing evidence class
 * 8) the geneticBackground class can be removed, since a simple string is not really sufficient - people can add the genetic background in comments.
 * 9) Phenotype should be moved to be a property of interaction, since a genetic interaction doesn't make sense without a phenotype specified. TODO: evaluate how PATO can be linked to BioPAX
 * 10) The position of Gene as a physicalEntity is not totally comfortable, since gene is a Schema (independent, abstract, continuant) according to the SOWA top level ontology (http://www.jfsowa.com/ontology/toplevel.htm), while DNA, RNA, protein and small molecule are objects (independent, physical, continuant) and complex is a Structure (mediating, physical, continuant) - both physical, while gene is abstract.  The exact position of gene needs some more thought.
 * 11) TODO: replace the gene definition with one from Sequence Ontology (SO).  See http://www.nature.com/nature/journal/v441/n7092/full/441398a.html
 * 12) OBI was evaluated on Jan.4.2007 for use in the evidence class, but was found to not be a good match at its current development stage. We should try it later. subversion repository revision 48  It would be useful to find out what use cases it is focused on, what the OBI team's requirements are for input of new terms that may be needed by pathway databases and level of development activity.

= Relation to external ontologies and plan for use =


 * 1) Controlled vocabularies for phenotype would be useful. While it is impossible to capture all phenotypes in a controlled vocabulary (they are effectively infinite), many common phenotypes, such as viability, are used in the study of genetic interactions and these could be captured.
 * 2) 'genetic experimental system' information can be stored in the existing EVIDENCE-CODE property in the geneticEvidence class. The PSI-MI controlled vocabulary that is already used to store experimental system types contains genetic interaction experimental system terms. An example system is "synthetic genetic analysis". - Anchor(expSystemCVNote)
 * 3) The following controlled vocabularies are available for describing phenotypes in various organisms (this is likely not an exhaustive list): - Anchor(phenotypeCVNote)
 * 4) * Suitable controlled vocabularies must be available in a relatively standard computable format (like OBO), must be referenceable (i.e. the terms must have unique identifiers) and must match the meaning of the property they will be used in.
 * 5) * A number of phenotype controlled vocabularies are available on the OBO web site in OBO format. The list as of mid-2006 is:
 * 6) ** Appear suitable for use in BioPAX:
 * 7) *** PATO - Currently under development, but the goal is to create a general phenotype ontology.
 * 8) *** cereal plant trait, used by Gramene
 * 9) *** mammalian phenotype, used by MGD and RGD
 * 10) *** human diseases, used by the NuGene project (the NIH controlled vocabularies are likely more complete)
 * 11) * Do not appear suitable for use in BioPAX:
 * 12) ** eVOC (Expressed Sequence Annotation for Humans) - limited set of human pathology terms
 * 13) ** medaka fish anatomy and development (under discussion)
 * 14) ** mouse pathology - limited set of phenoypes - main goal is to annotate images. The MGD ontology seems more complete.
 * 15) ** plant enviromental conditions - useful for decribing environmental perturbations, which this proposal doesn't cover.
 * 16) Other sources:
 * 17) * Appear suitable for use in BioPAX:
 * 18) ** The CYGD phenotype list contains a number of phenotype terms suitable for budding yeast, though may not be general enough because it doesn't seem to list lethality as a phenotype, just slow-growth.
 * 19) May be suitable for use in BioPAX:
 * 20) * The flybase query form displays a phenotype classification system, but it does not appear to be available as a separate, referenceable controlled vocabulary.
 * 21) Do not appear suitable for use in BioPAX:
 * 22) * http://www.phenomicdb.de/ is a database of phenotypes of multiple organisms. It contains general terms like lethal, though does not appear to organize them into a controlled vocabulary. Also, it relies on model organism databases as a data source.
 * 23) * Many GO biological process terms appear to be useful for representing phenotype, though it is not clear if these are suitable, since phenotype is the output of a biological process i.e. they are related, but do not mean the same thing.
 * 24) * ZFIN maintains a list of phenotype keywords for zebrafish, but they don't seem to have identifiers.
 * 25) Integration of the genetic interaction type and the gene perturbation types CVs into PSI-MI. - Anchor(PSINote)
 * 26) * PSI-MI interaction type contains 'genetic interaction', 'physical interaction' and 'colocalization' interaction terms. These types are suitable for use in the BioPAX 'interaction' class (in the INTERACTION-TYPE property). The genetic interaction term tree, as of mid-2006, is:




 * These seem to be taken from the MIPS evidence catalogue.
 * Some of the terms don't match the above classification: e.g. conditional synthetic lethal - is this conditional or synthetic? These don't overlap by the above classification.  Unless the condition term is describing another dimension of perturbation, like environmental perturbation (temperature, nutrition, like that term's children) instead of genetic perturbation.
 * Terms proposed here have been reported to the PSI-MI as two separate issues: experimental form and interaction type

= Future growth =


 * 1) The genetic interaction class could be subclassed into specific types of genetic interaction, though none of these currently would require additional properties or restrictions that aren't already in the geneticInteraction class.
 * 2) The Drees et al. system is comprehensive given normal background conditions, however there are other types of genetic interactions that depend on specific additional conditions. E.g. intergenic non-complementation requires more information about the complementation status of the interacting genes. Other similar interactions may be added in the future.
 * 3) Other perturbations, like environmental perturbation (e.g. temperature change from normal growth conditions, nutrition change), could be added either as a participant type or as experimental background.
 * 4) * This doesn't seem to be captured by many existing databases (though a controlled vocabulary does exist for plants).
 * 5) Describing the environmental conditions for an experiment (e.g. temperature, nutrition) is a useful piece of information in most genetic experiments, however it seems to be rarely captured by existing databases and it appears that it will need an extensive controlled vocabulary, which doesn't currently exist for many organisms. We can add a property to capture this information in the proposed geneticEvidence class in the future, once a controlled vocabulary is identified. E.g. new property: ENVIRONMENTAL-CONDITIONS in the geneticEvidence class. Note: this is different than environmental perturbation, which is a perturbation (deviation from normal), not a description of a background environmental condition e.g. temperature vs. temperature change.
 * 6) Detailed phenotype descriptions are difficult to provide, since the phenotype space is effectively infinite, though improved descriptions may be available in the future.
 * 7) * For instance, we may want to include descriptions specific to certain phenotypes e.g. rate of growth, shape of liver, etc.
 * 8) * Quantitative phenotypes are currently not well defined in biology. They are usually described in English (e.g. cell is red with punctate patches and an oval-like shape)  When these become more formally defined e.g. by existing databases, they could be added to the geneticEvidence class. The OBO phenotype group seems to be working on this.
 * 9) * Additional controlled vocabularies of phenotypes may be available in the future and these could be use in the PHENOTYPE property without further effort, though the documentation would need to be updated. We should always be searching for and supporting the creation of these controlled vocabularies for different organisms.
 * 10) The recording of multiple phenotypic outputs at the same time and their relation to each other as the result of the experimental perturbations is not well defined currently, however, this could be added in the future in the geneticEvidence class.
 * 11) More gene perturbations could be stored in the EXPERIMENTAL-FORM-TYPE property in geneticExperimentalForm e.g.
 * 12) * chemical-target: The gene has been specifically targeted by a small molecule known not to interfere with any other gene in the system
 * 13) ** This particular type of experimental form would benefit from describing the actual small molecule, which can be stored by adding a new property to the geneticExperimentalForm class.
 * 14) Additional  'genetic experimental system' terms, used in the existing EVIDENCE-CODE property in the geneticEvidence class, could be added to the PSI-MI controlled vocabulary. For example, a data provider may want to add a "tetrad analysis" term.
 * 15) Genetic background can be extended to describe a more structured description of the background by adding properties to the geneticBackground class.
 * 16) Chemical genetic interactions require their own type of interaction i.e. chemicalGeneticInteraction, which will be dealt with in a future proposal.
 * 17) Intergenic non-complementation interaction type could be added later. Intergenic non-complementation interaction is rare, but examples exist (e.g. sqt-1, sqt-2, sqt-3 and rol-8 combinations in C. elegans.
 * 18) The mathematical model that defines the genetic interaction, e.g. by statistical independance or inequality, could be stored.

= OWL considerations =


 * 1) Similar considerations as BioPAX Level 1 and 2.  All new classes should be disjoint from all sibling classes. (right-click on superclass in Protege and "set all subclasses disjoint".
 * 2) geneticEvidence is related to the existing evidence class e.g. they share a number of properties, but this is not specifically dealt with e.g. by stating that they are both subclasses of some more general evidence class. If this is necessary e.g. for reasoning, the general class could be created in the future without causing any backwards compatibility problems.

= Backward Compatibility =


 * 1) This proposal only adds classes and properties, thus is backwards compatible with Level 2.

= Notes =


 * 1) The gene class proposed here is equivalent to that proposed in the Gene Regulation Proposal - Anchor(geneNote)
 * 2) * Only gene databases should be referenced as unificationXrefs from gene instances. E.g. Entrez Gene or model organism databases.
 * 3) * The GENE-PRODUCT property of the gene class can be used when describing genetic interactions, but it is not required.
 * 4) Background: Gene mutations can perturb the system by altering the function of a gene. The mechanism is assumed to be perturbation of the resulting gene product structure in some way (many different ways are known). - Anchor(backgroundNote)
 * 5) Some genetic interactions are stored in GO using the IGI evidence code where the interactor is present in the "with" column.
 * 6) Because of the lack of suitable phenotype CVs for many model organisms, if a phenotype CV is not available, a user may create one, but must follow community best practices, such as from the OBO group.
 * 7) This proposal was significantly revised summer 2006 after feedback and discussions at the CSHL BioPAX meeting. Notes and implementation details from the original proposal, which will be useful for future expansion of this proposal, are at Genetic_Interactions/OldNotes

= References = Recommended reading:

Genetic interactions:
 * Systematic yeast synthetic lethal and synthetic dosage lethal screens identify genes required for chromosome segregation - PNAS 2005, Measday V, Baetz K, Guzzo J, Yuen K, Kwok T, Sheikh B, Ding H, Ueta R, Hoac T, Cheng B, Pot I, Tong A, Yamaguchi-Iwai Y, Boone C, Hieter P, Andrews B
 * Exploration of the function and organization of the yeast early secretory pathway through an epistatic miniarray profile - Cell 2005, Schuldiner M, Collins SR, Thompson NJ, Denic V, Bhamidipati A, Punna T, Ihmels J, Andrews B, Boone C, Greenblatt JF, Weissman JS, Krogan NJ
 * Derivation of genetic interaction networks from quantitative phenotype data - Genome Biology 2005, Drees BL, Thorsson V, Carter GW, Rives AW, Raymond MZ, Avila-Campillo I, Shannon P, Galitski T
 * The synthetic genetic interaction spectrum of essential genes - Nature Genetics 2005, Davierwala AP, Haynes J, Li Z, Brost RL, Robinson MD, Yu L, Mnaimneh S, Ding H, Zhu H, Chen Y, Cheng X, Brown GW, Boone C, Andrews BJ, Hughes TR
 * Global mapping of the yeast genetic interaction network - Science 2004, Tong AH, Lesage G, Bader GD, Ding H, Xu H, Xin X, Young J, Berriz GF, Brost RL, Chang M, Chen Y, Cheng X, Chua G, Friesen H, Goldberg DS, Haynes J, Humphries C, He G, Hussein S, Ke L, Krogan N, Li Z, Levinson JN, Lu H, Menard P, Munyana C, Parsons AB, Ryan O, Tonikian R, Roberts T, Sdicu AM, Shapiro J, Sheikh B, Suter B, Wong SL, Zhang LV, Zhu H, Burd CG, Munro S, Sander C, Rine J, Greenblatt J, Peter M, Bretscher A, Bell G, Roth FP, Brown GW, Andrews B, Bussey H, Boone C

Computational analysis of genetic interaction networks:
 * Combining biological networks to predict genetic interactions - PNAS 2004, Wong SL, Zhang LV, Tong AH, Li Z, Goldberg DS, King OD, Lesage G, Vidal M, Andrews B, Bussey H, Boone C, Roth FP
 * Systematic interpretation of genetic interactions using protein networks - Nature Biotechnology, Kelley R, Ideker T
 * Modular epistasis in yeast metabolism. - Nature Genetics 2005, Segre D, Deluna A, Church GM, Kishony R
 * Epistasis analysis with global transcriptional phenotypes - Nature Genetics 2005, Van Driessche N, Demsar J, Booth EO, Hill P, Juvan P, Zupan B, Kuspa A, Shaulsky G

More references are available at http://baderlab.org/MultiplePerturbationReferenceList

-

= Discuss this proposal =

Genetic_Interactions/Discussion this proposal.

= TODO =
 * It would be nice to have more figures. Class diagrams + diagrams of the interaction data in the worked examples.
 * It would be nice to provide alternate OWL file showing only proposed classes, with the original BioPAX Level 2 ontology imported. This does not seem to work in Protege 3.0 because properties from the imported ontology cannot be added to classes in the new ontology - even if both files use the same namespace.  If someone figures out how to get this to work, let me know.