Autism has a strong genetic basis, although the genetics of autism is complex and it is unclear whether ASD is explained more by multigene interactions or by rare mutations with major effects. Early studies of twins estimated the heritability of autism to be more than 90%; in other words, that genetics explains more than 90% of autism cases. This may be an overestimate; new twin data and models with structural genetic variation are needed. When only one identical twin is autistic, the other often has learning or social disabilities. For adult siblings, the risk of having one or more features of the broader autism phenotype might be as high as 30%, much higher than the risk in controls.

Genetic linkage analysis has been inconclusive; many association analyses have had inadequate power. For each autistic individual, mutations in more than one gene may be implicated. Mutations in different sets of genes may be involved in different autistic individuals. There may be significant interactions among mutations in several genes, or between the environment and mutated genes. By identifying genetic markers inherited with autism in family studies, numerous candidate genes have been located, most of which encode proteins involved in neural development and function. However, for most of the candidate genes, the actual mutations that increase the risk for autism have not been identified. Typically, autism cannot be traced to a Mendelian (single-gene) mutation or to single chromosome abnormalities such as fragile X syndrome or 22q13 deletion syndrome.

The large number of autistic individuals with unaffected family members may result from copy number variations (CNVs)—spontaneous alterations in the genetic material during meiosis that delete or duplicate genetic material. Sporadic (non-inherited) cases have been examined to identify candidate genetic loci involved in autism. Using array comparative genomic hybridization (array CGH), a technique for detecting CNVs, one study found them in 10% of families with one affected child. Some of the altered loci had been identified in previous studies of inherited autism; many were unique to the sporadic cases examined in this study. Hence, a substantial fraction of autism may be highly heritable but not inherited: that is, the mutation that causes the autism is not present in the parental genome. Although the fraction of autism traceable to a genetic cause may grow to 30–40% as the resolution of array CGH improves, several results in this area have been described incautiously, possibly misleading the public into thinking that a large proportion of autism is caused by CNVs and is detectable via array CGH, or that detecting CNVs is tantamount to a genetic diagnosis. The Autism Genome Project database contains genetic linkage and CNV data that connect autism to genetic loci and suggest that every human chromosome may be involved. It may be that using autism-related subphenotypes instead of the diagnosis of autism per se may be more useful in identifying susceptible loci.

Twin studies

Twin studies are a helpful tool in determining the heritability of disorders and human traits in general. They involve determining concordance of characteristics between identical (monozygotic or MZ) twins and between fraternal (dizygotic or DZ) twins. Possible problems of twin studies are: (1) errors in diagnosis of monozygocity, and (2) the assumption that social environment sharing by DZ twins is equivalent to that of MZ twins.

A condition that is environmentally caused without genetic involvement would yield a concordance for MZ twins equal to the concordance found for DZ twins. In contrast, a condition that is completely genetic in origin would theoretically yield a concordance of 100% for MZ pairs and usually much less for DZ pairs depending on factors such as the number of genes involved and assortative mating.

An example of a condition that appears to have very little if any genetic influence is irritable bowel syndrome (IBS), with a concordance of 28% vs. 27% for MZ and DZ pairs respectively. An example of a human characteristics that is extremely heritable is eye color, with a concordance of 98% for MZ pairs and 7–49% for DZ pairs depending on age.

Identical twin studies put autism's heritability in a range between 0.36 and 0.957, with concordance for a broader phenotype usually found at the higher end of the range. Autism concordance in siblings and fraternal twins is anywhere between 0 and 23.5%. This is more likely 2–4% for classic autism and 10–20% for a broader spectrum. Assuming a general-population prevalence of 0.1%, the risk of classic autism in siblings is 20- to 40-fold that of the general population.

Notable twin studies have attempted to shed light on the heritability of autism.

A small scale study in 1977 was the first of its kind to look into the heritability of autism. It involved 10 DZ and 11 MZ pairs in which at least one twin in each pair showed infantile autism. It found a concordance of 36% in MZ twins compared to 0% for DZ twins. Concordance of "cognitive abnormalities" was 82% in MZ pairs and 10% for DZ pairs. In 12 of the 17 pairs discordant for autism, a biological hazard was believed to be associated with the condition.

A 1979 case report discussed a pair of identical twins concordant for autism. The twins developed similarly until the age of 4, when one of them spontaneously improved. The other twin, who had suffered infrequent seizures, remained autistic. The report noted that genetic factors were not "all important" in the development of the twins.

In 1985, a study of twins enrolled with the UCLA Registry for Genetic Studies found a concordance of 95.7% for autism in 23 pairs of MZ twins, and 23.5% for 17 DZ twins.

In a 1989 study, Nordic countries were screened for cases of autism. Eleven pairs of MZ twins and 10 of DZ twins were examined. Concordance of autism was found to be 91% in MZ and 0% in DZ pairs. The concordances for "cognitive disorder" were 91% and 30% respectively. In most of the pairs discordant for autism, the autistic twin had more perinatal stress. A British twin sample was reexamined in 1995 and a 60% concordance was found for autism in MZ twins vs. 0% concordance for DZ. It also found 92% concordance for a broader spectrum in MZ vs. 10% for DZ. The study concluded that "obstetric hazards usually appear to be consequences of genetically influenced abnormal development, rather than independent aetiological factors.

A 1999 study looked at social cognitive skills in general-population children and adolescents. It found "poorer social cognition in males", and a heritability of 0.68 with higher genetic influence in younger twins.

In 2000, a study looked at reciprocal social behavior in general-population identical twins. It found a concordance of 73% for MZ, i.e. "highly heritable", and 37% for DZ pairs.

A 2004 study looked at 16 MZ twins and found a concordance of 43.75% for "strictly defined autism". Neuroanatomical differences (discordant cerebellar white and grey matter volumes) between discordant twins were found. The abstract notes that in previous studies 75% of the non-autistic twins displayed the broader phenotype.

Another 2004 study examined whether the characteristic symptoms of autism (impaired social interaction, communication deficits, and repetitive behaviors) show decreased variance of symptoms among monozygotic twins compared to siblings in a sample of 16 families. The study demonstrated significant aggregation of symptoms in twins. It also concluded that "the levels of clinical features seen in autism may be a result of mainly independent genetic traits.

An English twin study in 2006 found high heritability for autistic traits in a large group of 3,400 pairs of twins.

One critic of the pre-2006 twin studies said that they were too small and their results can be plausibly explained on non-genetic grounds.

Sibling studies

The importance of sibling studies lies in contrasting their results to those of fraternal (DZ) twin studies, plus their sample sizes can be much larger. Environment sharing by siblings is presumably different enough to that of DZ twins to shed some light on the magnitude of environmental influence. This should even be true to some extent regarding the prenatal environment. Unfortunately DZ twin study findings have yielded a very large range of variance and are error prone because of the apparent low concordance and the fact that they typically look at a small number of DZ pairs. For example, in studies involving 10 DZ pairs, a concordance below 10% would be impossible to determine precisely.

A study of 99 autistic probands which found a 2.9% concordance for autism in siblings, and between 12.4% and 20.4% concordance for a "lesser variant" of autism.

A study of 31 siblings of autistic children, 32 siblings of children with developmental delay, and 32 controls. It found that the siblings of autistic children, as a group, "showed superior spatial and verbal span, but a greater than expected number performed poorly on the set-shifting, planning, and verbal fluency tasks.

A 2005 Danish study looked at "data from the Danish Psychiatric Central Register and the Danish Civil Registration System to study some risk factors of autism, including place of birth, parental place of birth, parental age, family history of psychiatric disorders, and paternal identity." It found an overall prevalence rate of roughly 0.08%. Prevalence of autism in siblings of autistic children was found to be 1.76%. Prevalence of autism among siblings of children with Asperger's syndrome or PDD was found to be 1.04%. The risk was twice as high if the mother had been diagnosed with a psychiatric disorder. The study also found that "the risk of autism was associated with increasing degree of urbanisation of the child's place of birth and with increasing paternal, but not maternal, age.

A study in 2007 looked at a database containing pedigrees of 86 families with two or more autistic children and found that 42 of the third-born male children showed autistic symptoms, suggesting that parents had a 50% chance of passing on a mutation to their offspring. The mathematical models suggest that about 50% of autistic cases are caused by spontaneous mutations. The simplest model was to divide parents into two risk classes depending on whether the parent carries a pre-existing mutation that causes autism; it suggested that about a quarter of autistic children have inherited a copy number variation from their parents.

Other family studies

A 1994 looked at the personalities of parents of autistic children, using parents of children with Down's syndrome as controls. Using standardized tests it was found that parents of autistic children were "more aloof, untactful and unresponsive.

A 1997 study found higher rates of social and communication deficits and stereotyped behaviors in families with multiple-incidence autism. Autism was found to occur more often in families of physicists, engineers and scientists. Other studies have yielded similar results. Findings of this nature have led to the coinage of the term "geek syndrome".

A 2001 study of brothers and parents of autistic boys looked into the phenotype in terms of one current cognitive theory of autism. The study raised the possibility that the broader autism phenotype may include a "cognitive style" (weak central coherence) that can confer information-processing advantages.

A study in 2005 showed a positive correlation between repetitive behaviors in autistic individuals and obsessive-compulsive behaviors in parents. Another 2005 study focused on sub-threashold autistic traits in the general population. It found that correlation for social impairment or competence between parents and their children and between spouses is about 0.4.

A 2005 report examined the family psychiatric history of 58 subjects with Asperger's syndrome (AS) diagnosed according to DSM-IV criteria. Three (5%) had first-degree relatives with AS. Nine (19%) had a family history of schizophrenia. Thirty five (60%) had a family history of depression. Out of 64 siblings, 4 (6.25%) were diagnosed with AS.

Twinning risk

It has been suggested that the twinning process itself is a risk factor in the development of autism, presumably due to perinatal factors. However, three large-scale epidemiological studies have refuted this idea.

Proposed models

Twin and family studies show that autism is a highly heritable condition, but they have left many questions for researchers, most notably

• Why is fraternal twin concordance so low considering that identical twin concordance is high?
• Why are parents of autistic children typically non-autistic?
• Which factors could be involved in the failure to find a 100% concordance in identical twins?
• Is profound mental retardation a characteristic of the genotype or something totally independent?

Some researchers have speculated that what we currently refer to as "autism" may be a catch-all description for many yet unknown conditions with different genetic and/or environmental etiologies. This would appear to make the effort to find a genotype model a lot more difficult, and perhaps even pointless. Nevertheless, a number of genetic models have been proposed to try to explain the results of twin and sibling studies.

Mendelian

The original Mendelian model tried to explain observations using distinct genes existing in clearly dominant or recessive alleles. That would imply a recessive "autism gene" inherited from each of the parents. This kind of model is clearly too simple:

• It indicates that a sibling of an autistic individual should have 25% risk of having the autistic genotype, which is inconsistent with fraternal twin and sibling study results.
• It would require several characteristic features of autism to be caused by a single allele at a single locus.

Further considerations for the 'autism gene model' of also show contradictory implications:

• (a) only a small number of cases can be clearly linked to a possible genetic cause and these are often allele deletions;
• (b) the majority of patients with autism do not marry and do not have offspring which should result in a decreased incidence of the presumed gene in the general population.
• (c) the incidence of autism in the population has been increasing instead, making the likelihood of a single genetic cause extremely remote.

Mendel's later work and work based on it introduced polygenic inheritance, but taking into account linkage of genes required understanding where they were located - elucidating the role of the chromosomes.

Multigene

Reduced risk to relatives of probands and identical/fraternal twin ratios indicate that a multigene model is more likely to account for the autistic genotype. That is, at least two alleles would be involved, and most likely three to five. Researchers have suggested models of 15 and even up to 100 genes.

The fraternal twin results found by Ritvo et al (1985) and the broader phenotype results of Bolton et al (1994) suggest that a 2-gene model is plausible. Kolevzon et al (2004) proposed that the 3 characteristic symptoms of autism may be the result of 3 different alleles. Data supports the multiple-locus hypothesis and also that a 3-loci model is the best fit. Risch et al (1999) found results most compatible with a large number of loci (>= 15).

Given the significant prevalence of autism, perhaps 0.1% for classic autism and at least 0.6% for a broader spectrum, a multigene model has important implications. Since intelligence appears to be independent of the recognized characteristic symptoms of autism (and the diagnostic criteria) it is likely that many individuals are very autistic yet highly functional, allowing them to escape a diagnosis altogether. So the prevalence of the autistic genotype may be considerably higher than thought. And if multiple alleles are part of the genotype, then each allele must have relatively high prevalence in the general population.

Two family types

In this model most families fall into two types: in the majority, sons have a low risk of autism, but in a small minority their risk is near 50%. In the low-risk families, sporadic autism is mainly caused by spontaneous mutation with poor penetrance in daughters and high penetrance in sons. The high-risk families come from (mostly female) children who carry a new causative mutation but are unaffected and transmit the dominant mutation to grandchildren.

Epigenetic

Several epigenetic models of autism have been proposed. These are suggested by the occurrence of autism in individuals with fragile X syndrome, which arises from epigenetic mutations, and with Rett syndrome, which involves epigenetic regulatory factors. An epigenetic model would help explain why standard genetic screening strategies have so much difficulty with autism.

Genomic imprinting

Genomic imprinting models have been proposed; one of their strengths is explaining the high male-to-female ratio in ASD. One hypothesis is that autism is in some sense diametrically opposite to schizophrenia and other psychotic-spectrum conditions, that alterations of genomic imprinting help to mediate the development of these two sets of conditions, and that ASD involves increased effects of paternally expressed genes, which regulate overgrowth in the brain, whereas schizophrenia involves maternally expressed genes and undergrowth.

Environmental interactions

Though autism's genetic factors explain most of autism risk, they do not explain all of it. A common hypothesis is that autism is caused by the interaction of a genetic predisposition and an early environmental insult. Several theories based on environmental factors have been proposed to address the remaining risk. Some of these theories focus on prenatal environmental factors, such as agents that cause birth defects; others focus on the environment after birth, such as children's diets. All known teratogens (agents that cause birth defects) related to the risk of autism appear to act during the first eight weeks from conception, strong evidence that autism arises very early in development. Although evidence for other environmental causes is anecdotal and has not been confirmed by reliable studies, extensive searches are underway.

Candidate gene loci

A number of alleles have been shown to have strong linkage to the autism phenotype. In many cases the findings are inconclusive, with some studies showing no linkage. Alleles linked so far strongly support the assertion that there is a large number of genotypes that are manifested as the autism phenotype. At least some of the alleles associated with autism are fairly prevalent in the general population, which indicates they are not rare pathogenic mutations. This also presents some challenges in identifying all the rare allele combinations involved in the etiology of autism.

Primary

Gene	Locus	Description
?	16p11.2	A 2008 study observed a de novo deletion of 593 kb on this chromosome in about 1% of persons with autism, and similarly for the reciprocal duplication of the region. Another 2008 study also found duplications and deletions associated with ASD at this locus.
SERT (SLC6A4)	17q11.2	This gene locus has been associated with rigid-compulsive behaviors. Notably, it has also been associated with depression but only as a result of social adversity, although other studies have found no link. Significant linkage in families with only affected males has been shown. Researchers have also suggested that the gene contributes to hyperserotonemia. However, a 2008 meta-analysis of family- and population-based studies found no significant overall association between autism and either the promoter insertion/deletion (5-HTTLPR) or the intron 2 VNTR (STin2 VNTR) polymorphisms.
GABRB3, GABRA4	multiple	GABA is the primary inhibitory neurotransmitter of the human brain. Ma et al (2005) concluded that GABRA4 is involved in the etiology of autism, and that it potentially increases autism risk through interaction with GABRB1. The GABRB3 gene has been associated with savant skills. The GABRB3 gene deficient mouse has been proposed as a model of ASD.
Engrailed 2 (EN2)	7q36.2	Engrailed 2 is believed to be associated with cerebellar development. Benayed et al (2005) estimate that this gene contributes to as many as 40% of ASD cases, about twice the prevalence of the general population. But at least one study has found no association.
?	3q25-27	A number of studies have shown a significant linkage of autism and Asperger's syndrome with this locus. The most prominent markers are in the vicinity of D3S3715 and D3S3037.
Reelin	7q21-q36	In adults, Reelin glycoprotein is believed to be involved in memory formation, neurotransmission, and synaptic plasticity. A number of studies have shown an association between the REELIN gene and autism, but a couple of studies were unable to duplicate linkage findings.
SLC25A12	2q31	This gene encodes the mitochondrial aspartate/glutamate carrier (AGC1). It has been found to have a significant linkage to autism in some studies, but linkage was not replicated in others, and a 2007 study found no compelling evidence of an association of any mitochondrial haplogroup in autism.
HOXA1 and HOXB1	multiple	A link has been found between HOX genes and the development of the embryonic brain stem. In particular, two genes, HOXA1 and HOXB1, in transgenic 'knockout' mice, engineered so that these genes were absent from the genomes of the mice in question, exhibited very specific brain stem developmental differences from the norm, which were directly comparable to the brain stem differences discovered in a human brain stem originating from a diagnosed autistic patient. Conciatori et al (2004) found an association of HOXA1 with increased head circumference. A number of studies have found no association with autism. The possibility remains that single allelic variants of the HOXA1 gene are insufficient alone to trigger the developmental events in the embryo now associated with autistic spectrum conditions. Tischfield et al published a paper which suggests that because HOXA1 is implicated in a wide range of developmental mechanisms, a model involving multiple allelic variants of HOXA1 in particular may provide useful insights into the heritability mechanisms involved. Additionally, Ingram et al alighted upon additional possibilities in this arena. Transgenic mouse studies indicate that there is redundancy spread across HOX genes that complicate the issue, and that complex interactions between these genes could play a role in determining whether or not a person inheriting the requisite combinations manifests an autistic spectrum condition—transgenic mice with mutations in both HOXA1 and HOXB1 exhibit far more profound developmental anomalies than those in which only one of the genes differs from the conserved 'norm'. In Rodier's original work, teratogens are considered to play a part in addition, and that the possibility remains open for a range of teratogens to interact with the mechanisms controlled by these genes unfavourably (this has already been demonstrated using valproic acid, a known teratogen, in the mouse model).
PRKCB1	16p11.2	Philippi et al (2005) found a strong association between this gene and autism. This is a recent finding that needs to be replicated.
FOXP2	7q31	The FOXP2 gene is of interest because it is known to be associated with developmental language and speech deficits. An association to autism appears to be elusive, nonetheless.
UBE3A	15q11.2–q13	The maternally expressed imprinted gene UBE3A has been associated with Angelman syndrome. MeCP2 deficiency results in reduced expression of UBE3A in some studies.
Shank3 (ProSAP2)	22q13	The gene called SHANK3 (also designated ProSAP2) regulates the structural organization of neurotransmitter receptors in post-synaptic dendritic spines making it a key element in chemical binding crucial to nerve cell communication. SHANK3 is also a binding partner of chromosome 22q13 (i.e. a specific section of Chromosome 22) and neuroligin proteins; deletions and mutations of SHANK3, 22q13 (i.e. a specific section of Chromosome 22) and genes encoding neuroligins have been found in some people with autism spectrum disorders. Mutations in the SHANK3 gene have been strongly associated with the autism spectrum disorders. If the SHANK3 gene is not adequately passed to a child from the parent (haploinsufficiency) there will possibly be significant neurological changes that are associated with yet another gene, 22q13, which interacts with SHANK3. Alteration or deletion of either will effect changes in the other. A deletion of a single copy of a gene on chromosome 22q13 has been correlated with global developmental delay, severely delayed speech or social communication disorders and moderate to profound delay of cognitive abilities. Behavior is described as "autistic-like" and includes high tolerance to pain and habitual chewing or mouthing (see also 22q13 deletion syndrome). This appears to be connected to the fact that signal transmission between nerve cells is altered with the absence of 22q13. SHANK3 proteins also interact with neuroligins at the synapses of the brain further complicating the widespread effects of changes at the genetic level and beyond.
NLGN3	Xq13	Neuroligin is a cell surface protein (homologous to acetylcholinesterase and other esterases) that binds to synaptic membranes. Neuroligins organize postsynaptic membranes that function to transmit nerve cell messages (excitatory) and stop those transmissions (inhibitory); In this way, neuroligins help to ensure signal transitions between nerve cells. Neuroligins are also regulate the maturation of synapses and ensure there are sufficient receptor proteins on the synaptic membrane. Mice with a neuroligin-3 mutation exhibit poor social skills but increased intelligence. Though not present in all individuals with autism, these mutations hold potential to illustrate some of the genetic components of spectrum disorders. However, a 2008 study found no evidence for involvement of neuroligin-3 and neuroligin-4x with high-functioning ASD.
MET	7q31	The MET gene (MET receptor tyrosine kinase gene) linked to brain development, regulation of the immune system, and repair of the gastrointestinal system, has been linked to autism. This MET gene codes for a protein that relays signals that turn on a cell’s internal machinery. Impairing the receptor’s signaling interferes with neuron migration and disrupts neuronal growth in the cerebral cortex and similarly shrinks the cerebellum—abnormalities also seen in autism. It is also known to play a key role in both normal and abnormal development, such as cancer metastases (hence the name MET). A mutation of the gene, rendering it less active, has been found to be common amongst children with autism. Mutation in the MET gene demonstrably raises risk of autism by 2.27 times.
neurexin 1	2q32	In February 2007, researchers in the Autism Genome Project (an international research team composed of 137 scientists in 50 institutions) reported possible implications in aberrations of a brain-development gene called neurexin 1 as a cause of some cases of autism. Linkage analysis was performed on DNA from 1,181 families in what was the largest-scale genome scan conducted in autism research at the time. The objective of the study was to locate specific brain cells involved in autism to find regions in the genome linked to autism susceptibility genes. The focus of the research was copy number variations (CNVs), extra or missing parts of genes. Each person does not actually have just an exact copy of genes from each parent. Each person also has occasional multiple copies of one or more genes or some genes are missing altogether. The research team attempted to locate CNVs when they scanned the DNA. Neurexin 1 is one of the genes that may be involved in communication between nerve cells (neurons). Neurexin 1 and other genes like it are very important in determining how the brain is connected from cell to cell, and in the chemical transmission of information between nerve cells. These genes are particularly active very early in brain development, either in utero or in the first months or couple of years of life. In some families their autistic child had only one copy of the neurexin 1 gene. Besides actually locating yet another possible genetic influence (the findings were statistically insignificant), the research also reinforced the theory that autism involves many forms of genetic variations. A 2008 study implicated the neurexin 1 gene in two independent subjects with ASD, and suggested that subtle changes to the gene might contribute to susceptibility to ASD.
CNTNAP2	7q35-q36	Multiple 2008 studies have identified a series of functional variants in the CNTNAP2 gene, a member of the neurexin superfamily, that implicate it as contributing to autism.
GSTP1	11q13	A 2007 study suggested that the GSTP1*A haplotype of the glutathione S-transferase P1 gene (GSTP1) acts in the mother during pregnancy and increases the likelihood of autism in the child.
PRL, PRLR, OXTR	multiple	A 2008 study found preliminary data supporting the hypothesis that ASD is associated with allelic variants of genes needed for typical affiliative behaviors. The strongest results were obtained for the PRL, PRLR, and OXTR genes.

Others

There is a large number of other candidate loci which either should be looked at or have been shown to be promising. Several genome-wide scans have been performed identifying markers across many chromosomes.

A few examples of loci that have been studied are the 17q21 region , the 3p24-26 locus, PTEN, and 15q11.2–q13.

Homozygosity mapping in pedigrees with shared ancestry and autism incidence has recently implicated the following candidate genes: PCDH10, DIA1 (formerly known as C3ORF58), NHE9, CNTN3, SCN7A, and RNF8. Several of these genes appeared to be targets of MEF2, one of the transcription factors known to be regulated by neuronal activity and that itself has also recently been implicated as an autism-related disorder candidate gene.

Other possible candidates include:

• SLC6A2 (Social phobia)
• FMR1 (Fragile-X)
• 5-HT-1Dbeta (OCD)
• 7q11.23 (William's syndrome, language impairment)
• 4q34-35, 5q35.2-35.3, 17q25 (Tourette syndrome)
• 2q24.1-31.1 (Intelligence)
• 6p25.3-22.3 (Verbal IQ)
• 22q11.2 (Visio-Spatial IQ)

References

Further reading

• O'Roak BJ, State MW (2008). "Autism genetics: strategies, challenges, and opportunities". Autism Res 1 (1): 4–17. An excellent summary of the state and possible future directions of autism genetics research as of late 2007.
• Fisch GS (2008). "Syndromes and epistemology II: is autism a polygenic disorder?". Am J Med Genet A Advocates common disease rare allele (CDRA) over polygenic approaches.
• Losh M, Sullivan PF, Trembath D, Piven J (2008). "Current developments in the genetics of autism: from phenome to genome". J Neuropathol Exp Neurol Focuses on attempts to match genes to behavior.

External links

• Autism Genetic Resource Exchange (AGRE) - 'the world's first collaborative gene bank for autism'