Gene
Intersection Method Example One: Tourette’s Related Conditions
L. Van Warren MS CS, AE wdv.com
|
|||||||||||||||||||||||||||||||||||||||||
Introduction
Sometimes a sentence read in passing can connect two islands of knowledge to enable deeper understanding. If cellular knowledge is a network of interconnected facts, the definition is recursive. At the root, the islands of knowledge are things such as genes, proteins, and cellular processes.
|
|||||||||||||||||||||||||||||||||||||||||
For the last decade, the ubiquity of information is changing the practices we use to gain understanding.1 The shotgun sequencing method that enabled the human genome project was initially criticized as a “fishing expedition”.2 Combinatorial chemistry techniques have suffered similar disdain. We live in the first generation of “hot and cold running knowledge”. Those of us in this transition are rapidly adopting new thinking practices that exploit the explosion of genomic knowledge.
The decisive moment comes at the instant that knowledge produces benefit for the individual. Knowledge, like all information, is functionally useless until the moment that it is applied.
In previous work, a theory of knowledge mapping was developed to visualize complex relationships in biological systems.3 This article describes a computational algorithm for finding genes common to related conditions. It is called, “Gene Intersection Method” or GIM for short. GIM can be used to design gene chips for clinical laboratory assessment. GIM prioritizes results such that genes of greatest influence are ranked highest.
History Repeats Itself
In medieval times, physical illness was explained in vague and unsatisfactory terms. Metabolism and bodily function were seen to depend on elemental influences such as earth, wind, fire, and water. A complete dissection of the living organism, and its companion sciences required several hundred years and is just beginning at the genomic level. An explosion in factual knowledge has continued. This knowledge often contradicts intuitive explanations and traditions.
In popular culture, disorders of mind have suffered an obstruction in understanding similar to those of somatic illness of times past. Neuron biology and imaging techniques such as functional MRI, PET‑scanning, and neural circuit mapping are changed this rapidly and profoundly.4 Brain pathologies at several decades of scale are being understood from a functional point of view. Molecular biologists, psychologists, and computer scientists are working from the bottom up enabling a new age of genomic understanding.
Disorders of Mind vs. Body
The terms, “mental illness”, and “psychiatric illness” conjure images of hopelessness and incurability. Gene-based assessment offers the possibility of better diagnosis and treatments. 5 Treatments based on these have the potential to correct neurochemical disequilibrium of genetic origin. The first application of the gene intersection method will be for disorders of the mind. In this, they demonstrate significant opportunity. |
|||||||||||||||||||||||||||||||||||||||||
Morphology vs. Biochemistry
Consider two views of the functional brain. One is the macroscopic architecture, the gross anatomy of neural circuitry.6 Obsessive-compulsive disorder (OCD) is often referenced as a “basal ganglia disorder”. 7 The basal ganglia are involved in a constant neuronal feedback loops that enable initiation of new tasks and maintenance of current tasks, for example holding a glass of wine while talking. 6
Basal Ganglia – Macroscopically8
The other view is from the nanoscopic level of synaptic transmission via neurotransmitters and receptors Complex gene-expression networks modulate an elegant balance involving multiple players on multiple tiers of regulation. In the small, tissue specific receptors and multiple neurotransmitters are regulated in magnificent orchestration. Fixing one’s gaze or holding a glass are amazing feats of biological control.
When players in this neurochemical orchestra are absent, broken, or out of tune, the result can be chaos for equilibrium of the mind. Attempts to change the level of one receptor or neurotransmitter affect a complex and nonlinear regulatory network. Because of the complexity of this network, it is necessary to determine all receptor and ligand players, not just for the idealized person, but for the afflicted individual. It is essential to measure what the regulator network is supposed to be doing versus what the network is actually doing. Adding to the mix is the dynamic and chaotic behavior of biological systems, especially the mind. Did you get enough sleep last night? |
|||||||||||||||||||||||||||||||||||||||||
So we have two views. The first view is a top down architectural view whose exploration begins macroscopically. The second is a bottom up biochemical building block view. The root of the latter view is the gene expression level.
With both views, the “highway map” of neurochemical communication is being understood, enabling better assessment of disorders of mind.
Serotonin Highway – Simplified Macroscopic View10
Discussion of the First Example
At the clinical level, several disorders of mind appear to be functionally related. These disorders are Tourette’s syndrome, depression, obsessive-compulsive disorder (OCD), attention deficit-hyperactivity disorder (ADHD), tics, and sleep disorders.11 Recent advances in gene ontology enable discovering the genes common to related conditions. This set of disorders seemed a good test case for gene intersection method.
GIM discovers genes common to related conditions. The method provides a mathematical definition of related conditions. It enables identification of disorders that are related by involvement of a common gene. Gene intersection method speeds the definition and construction of diagnostic and monitoring tests. The presumption is that accurate genetic diagnosis will result in treatments that are more effective. Eventually real-time monitoring will be enabled knowing what to monitor for specifically afflicted individuals, or improving quality of life for normal individuals. |
|||||||||||||||||||||||||||||||||||||||||
Leckman and Cohen’s excellent treatise on Tourette’s Syndrome opens with the following definition: 11
Tourette’s syndrome is a developmental, neuropsychiatric disorder defined by persistent, motor and vocal tics, […] obsessions, compulsions, and attentional difficulties. Often based upon a multigenerational, genetic predisposition…
They then go on to list eleven related conditions that are explored below using the gene intersection method and filtering the results to include only those entries for which gene description and locus is known.
Let us begin by using GIM to find genes common to pairs of disorders.
First Results of Gene Intersection (GIM) Method
|
|||||||||||||||||||||||||||||||||||||||||
Compounded GIM: Genes Common to Triplets of Related Disorders
The method can be compounded and used to find genes shared by three conditions and so on. These provide interesting and pivotal genes to focus on. Note that fewer genes are shared among progressively longer lists of disorders. List compression occurs as non-common genes are eliminated. This offsets combinatorial complexity and leaves us focusing on the genes of most concern first. We will boil down Tourette’s related conditions to a few special genes listed below:
The following table is the four way case. This small table is the most important one for now.
|
|||||||||||||||||||||||||||||||||||||||||
A Brief Digression on Equivalence Relationships
Set theory12 teaches three kinds of relationships these table entries can have, easily remembered using the mnemonic SRT. S stands for symmetric. A relationship is symmetric if:
(a isRelatedTo b) implies ( b isRelatedTo A)
All the relationships we are considering here are symmetric. If a gene is common to bipolar disorder and OCD, then the reverse is true; It is common to OCD and bipolar disorder. This distinction may sound ridiculous at first but it is mathematically necessary. The R stands for reflexive. A relationship is reflexive if it is true for itself. In the first table, there are 90 genes found for anxiety. The reflexive case is just the number of genes common to anxiety and anxiety. This number is a measure of the complexity of the disorder and research progress in the disorder. This number increases over time until all the genes relating to that disorder have been discovered and verified. All numbers along the principal diagonal of the table are reflexive entries. So we see that 90 genes have been discovered that relate to anxiety disorders.
The T in SRT stands for transitive. A relationship is transitive if:
(a isRelatedTo b) AND (b isRelatedTo c) implies (a isRelatedTo c)
Incidentally, the relationship “equals” satisfies all three of SRT.
It is informative to consider everyday relationships which do, or do not possess these properties, but I digress. The intent of the graphic is this; Anxiety disorders share nine “genes” with bipolar disorders. That is, of the 90 genes associated with anxiety and the 63 genes associated with bipolar disorder, there are nine genes in common between the two conditions. As a matter of parity (90 + 63) – 9 = 144, genes are not shared, a measure of the relative complexity of the two illnesses and the effort that has been expended to address them. We would like to understand the neurochemical pathways that the genes share. Doing so gives us the value of understanding two disorders for the price of one. It allows us to prioritize our concerns.
A Strength of Genetic Intersection Method
A strength of GIM is the identification of hotspots shared among disorders. These constitute the “low-hanging fruit” for clinical diagnosis and treatment. These are the gene landmarks in neurochemical pathways. In the graphic above, those conditions that share genes are sorted and colored.
A Weakness of the GIM Method
In some pairs of disorders there appears to be no common genetic ground, no shared genes. That does not mean there aren’t, it only means that we have not yet discovered them if they happen to exist. This is a weakness of the GIM method. Emphasizing this point: Zero common genes do not mean none exists, only that we haven’t found them yet. Thus, GIM lives downstream of gene discovery. A potential source of false-positives occurs when a gene has a false-association history. In the worst case, one would include genes on a diagnostic gene chip that are not relevant and do not need to be monitored. |
|||||||||||||||||||||||||||||||||||||||||
Explanation of GIM Method: Step One
In step one, the terms, and keywords most frequently associated with a set of related conditions in a complex disease are identified. There is a certain amount of lexical art when it comes to term selection. Some terms such as “depression” seem ambiguous. To avoid vague results terms should pertain specifically to the condition of interest. Sometimes casting a wider net with a more general term can be beneficial. In our case the search terms are:
Step Two: Execute the Search Using the Search Terms in OMIM
Entering a search terms in OMIM13 produces a list of candidate “genes”. (For the naïve method, any OMIM entry is considered a “gene”). Take the term “ADHD” for example. OMIM returns 20 entries, each with three fields, the OMIM entry number, the name of the entry, and gene locus as follows:
#143465 ATTENTION DEFICIT-HYPERACTIVITY DISORDER; ADHD Gene map locus 17p11, 16p13, 6q12, 5p13, 5p15.3, 4p16.1-p15.3 *126452 DOPAMINE RECEPTOR D4; DRD4 Gene map locus 11p15.5 *126455 SOLUTE CARRIER FAMILY 6 (NEUROTRANSMITTER TRANSPORTER, DOPAMINE), MEMBER 3; SLC6A3 Gene map locus 5p15.3 *104210 ALPHA-2A-ADRENERGIC RECEPTOR; ADRA2A Gene map locus 10q24-q26 #192430 VELOCARDIOFACIAL SYNDROME Gene map locus 22q11.2 +126453 DOPAMINE RECEPTOR D5; DRD5 DYSTONIA, PRIMARY CERVICAL, INCLUDED Gene map locus 4p16.1-p15.3 #137580 GILLES DE LA TOURETTE SYNDROME; GTS CHRONIC MOTOR TICS, INCLUDED Gene map locus 11q23 #600202 DYSLEXIA, SUSCEPTIBILITY TO, 2; DYX2 Gene map locus 6p22.2 +305400 FACIOGENITAL DYSPLASIA FGD1 GENE, INCLUDED; FGD1, INCLUDED Gene map locus Xp11.21 *609678 SLIT- AND NTRK-LIKE FAMILY, MEMBER 1; SLITRK1 %608904 ATTENTION DEFICIT-HYPERACTIVITY DISORDER, SUSCEPTIBILITY TO, 2 Gene map locus 17p11 %608903 ATTENTION DEFICIT-HYPERACTIVITY DISORDER, SUSCEPTIBILITY TO, 1 Gene map locus 16p13 *608396 SOLUTE CARRIER FAMILY 9 (SODIUM/HYDROGEN EXCHANGER), ISOFORM A9; SLC9A9 Gene map locus 3q21 %608906 ATTENTION DEFICIT-HYPERACTIVITY DISORDER, SUSCEPTIBILITY TO, 4 Gene map locus 5p13 %608905 ATTENTION DEFICIT-HYPERACTIVITY DISORDER, SUSCEPTIBILITY TO, 3 Gene map locus 6q12 %608631 ASPERGER SYNDROME, SUSCEPTIBILITY TO, 2 Gene map locus 17p13 #605899 GLYCINE ENCEPHALOPATHY; GCE HYPERGLYCINEMIA, TRANSIENT NEONATAL, INCLUDED; TNH, INCLUDED Gene map locus 16q24, 9p22, 3p21.2-p21.1 +162200 NEUROFIBROMATOSIS, TYPE I; NF1 NEUROFIBROMIN, INCLUDED Gene map locus 2p22-p21, 17q11.2 %102300 RESTLESS LEGS SYNDROME 1 Gene map locus 12q12-q21 *191290 TYROSINE HYDROXYLASE; TH Gene map locus 11p15.5 |
|||||||||||||||||||||||||||||||||||||||||
The summary result is as follows:
Step Three: Processing the Gene List For Each Search Term
As a positive control, the genes of each search term are compared with itself. If the software is functioning properly this trivially produces the original gene count for the search term. This is mathematically similar to dotting the gene vector returned by a search term with itself. It just gives the length, that is, the number of “genes” for a given disorder. In what follows the term “vector” and “list” will be used interchangeably. |
|||||||||||||||||||||||||||||||||||||||||
To accomplish this means performing some necessary but mundane text processing. We must construct the result vector from the raw result by stripping everything except the gene ID field. The symbols in front of each gene ID have specific meanings given in the following table. OMIM returns a single character prefix that indicates the nature of the entry. This is expanded into a slightly more mnemonic two-character prefix. The first character pertains to phenotype, the expressed trait, the second character of the prefix pertains to genotype, or possible trait. P in upper case indicates phenotype is known and described. G in upper case indicates genotype locus is known. Lower case p indicates phenotype is described with unknown locus.
The recoded result is more readable, and eliminates problems with symbols that have special meaning to the text processing software. Each list returned from a term search is stored in a file. The file suffix is initially .txt to indicate a text file. The file suffix is changed in the process to prevent stepping on the original data. This allows supervision of the process until it is verified. |
|||||||||||||||||||||||||||||||||||||||||
Step Four: Exhaustive Search For Common Terms
In this step, every gene list is compared with every other gene list for each of the 11 search terms that produced them. This requires 114=14,641 file comparisons for the four-way case, but null files are not compared, reducing the search cost significantly. The results are formatted in tabular form. The table is sorted by the number of common genes and colored to highlight conditions which share genes. The intensity of the color indicates the degree of overlap between the conditions. The main diagonal of the original matrix is a useful navigational landmark. The dot product principle applies to higher dimensional matrices because it remains a one-dimensional vector. This is critical for the higher order cases.
Step Five: Overlapping Genes are Inspected For Significance
A count indicating the number of common genes common is produced by the software, as well as a list of the actual genes. This partitions the results into two groups; Entries whose locus and phenotype are well characterized and those which are not. Take for example the genes shared by anxiety and bipolar disorder in the small table below. The last entry will be discarded as no gene locus is associated with it.
|
|||||||||||||||||||||||||||||||||||||||||
Here we can use the prefix codes. We are most interested in those conditions for which we can establish a gene chip normal baseline and mutated cases. This will be true for entries prefixed with PG, meaning phenotype and genotype are well characterized, and for entries prefixed with _G, meaning some gene association with this disorder exists in the literature. Entries prefixed with P_ and p_ are discarded, since no gene specific information is available yet for these disorders. The remaining gene ID’s are retained.
Author’s Note: It was with excitement that the names of genes shared by anxiety and bipolar disorder were retrieved. One often reaches for but rarely experiences that sense of progress. It is enriching to see precursors, degradation enzymes and NT transport systems implicated as well as long-known NT. The similarity between hormones and neurotransmitters is striking. Long distance and short-range cell-to-cell signaling share mechanism, extending the boundary of the “nervous system”. Hormones do their work binding to free cytosolic receptors which dimerize, enter the nucleus, and bind to DNA and upregulate transcription. Neurotransmitters bind to transmembrane receptors facilitating long and short-term state changes in cell state. In some sense, the synapse itself looks cytosolic. The difference being that hormones are ligands for transcription factors while neurotransmitters enable ion cascades. Hormones can trigger growth cascades which also have ionic involvement, for example the role of calcium ion concentration in nuclear envelope breakdown.
The take home lesson is that we must understand all players in the pathways these genes share. Compensation of the entire pathway will reduce disorders of the mind. Single factor solutions have problems due to the long delay between dose changes and assessment. 14
We must pull these sweater threads to look for other gene loci with similar roles. This provides handles in the control of neurochemical equilibrium. |
|||||||||||||||||||||||||||||||||||||||||
Two Clinical Approaches
Two clinical approaches are suggested:
Procedure 1: Macro to Micro - Clinically Driven Approach
In this approach, we move from clinical diagnosis to gene identification to gene tailored treatment. This is especially useful when a constellation of symptoms is present, each of symptoms can be used to indicate a screen to reveal which genes are involved. When specific functional genes are identified, therapies that are targeted to the specific gene products (or lack of!) can be performed.
The patient is screened with one or more gene chips to determine if the candidate genes are in fact mutated. This confirms and verifies the clinical assessment and diagnosis, providing a biochemical "second opinion" and confirming or refuting diagnostic theories being applied to a specific patient. A powerful advantage is that this works whether the patient is being seen for the first time, or has been in long-term treatment.
Procedure 2: Micro to Macro - Gene Driven Approach:
When patients are referred into a justice or psychiatric system without a solid history, they can be screened and eventually treated for a broad range conditions by looking directly at the results of gene chip assays. This enables understanding of the forces affecting them and faster routing to effective treatment. Many of the privacy and sampling issues relating to DNA have already been defined, so this is becoming more practical.
This kind of custom tailored genetic profile will also allow for ideal patient follow up to occur and life-damaging or threatening events to be predicted, assessed, and avoided.
Once genetic confirmation is achieved, it is not necessary to retest frequently. Annual or bi-annual retesting should be done, since genetic methods are evolving rapidly and this process is becoming more comprehensive. The entire diagnostic procedure should be repeated as improvements in the gene databases occur. At this writing, six months is probably adequate for running the deductive diagnostic software, unless new gene functions are discovered and announced. The range of conditions addressed can be broadened as well. One could allow clinicians to enter a variety of patient symptoms and come up with a "custom profile", for example asthma and anxiety.
Conclusions
In summary, it is possible to work in either direction; from clinical condition to phenotype to genotype and visa versa. The goal is personally tailored treatement. A bioinformatics method of finding genes common to related disorders has been described and tested. This method makes possible the definition of custom gene chips using for screening disorders of the mind in the eleven conditions reviewed above. A potential schematic is given below.
|
|||||||||||||||||||||||||||||||||||||||||
Acknowledgments Support of this work by Lynn Mittelstaedt Warren, family and friends is gratefully acknowledged. The author would like to express deep appreciation regarding the continuing excellence of NCBI, OMIM, and PubMed data repositories. These are the great national resources of the genomic age. Thanks to Dr. Steven J. Fliesler of St. Louis University School of Medicine and Dr. Christopher Lamps of UAMS College of Medicine for their encouragement.
Abbreviations OMIM: Online Mendelian Inheritance In Man NCBI: National Center for Biotechnology Information PubMed: National Library of Medicine Division sSRI: Selective Serotonin Reuptake Inhibitor NT: Neurotransmitter or Neurotransmitters
References
1. Lynch, Clifford: "Searching the Internet," Scientific American, March 1997 2. Service, Robert F.; Big Biology Is Bad Biology; Science: Vol. 291. no. 5507, p. 1182, 16 February 2001 3. Warren, L. Van: Knowledge Mapping the Corpus; www.wdv.com Feb 23, 2002 4. Cabeza, Roberto and Nyberg, Lars; Imaging Cognition II: An Empirical Review of 275 PET and fMRI Studies; Journal of Cognitive Neuroscience. Vol. 12:1-47, 2000 5. Perou, Charles M. et al; Molecular Portraits of Breast Cancer; Nature 406, 747-752, (17 August 2000 6. Best, Ben; The Basal Ganglia, Chapter 2 – Gross Neuroanatomy Web Page 1990-2006 7. Osborn, Ian; Tormenting Thoughts and Secret Rituals – Dell, Random House ISBN 0-440-50847-9 (1998) 8. Kalat, James; Biological Psychology 8th edition– Wadsworth Thomson Publishing 2001 9. Lundbeck Institute; Brain Atlas - Neurotransmitters; www.brainexplorer.org (2005) 10. University of Utah; Genetic Science Learning Center; Beyond the Reward Pathway (2006) 11. Leckman, James F. and Cohen, Donald J.; Tourette’s Syndrome – Tics, Obsessions, Compulsions; John Wiley & Sons (1999) ISBN – 0-471-16037-7 12. Wikipedia; Equivalence Relation; Feb 14, 2006 13. OMIM; Online Mendelian Inheritance in Man; National Center for Biotechnology Information (2006) 14. Mayberg, Helen S. et al; Regional Metabolic Effects of Fluoxetine in Major Depression: Serial Changes and Relationship to Clinical Response; Biological Psychiatry (2000) Vol 48. p 831
Author Information This theme of understanding complex systems has been the key controlling idea in two previous theses - one in computer science dealing with spatial decomposition of complex scenes and one in engineering dealing with the ontology and simulation of complex structural systems. A more complete bio is posted on wdv.com.
last updated June 21, 2006
|
