Progress in the genetics of autosomal recessive nonsyndromic mental retardation (ARNSMR)

Volkmar Weiss

Home

IQ Genetics by Data Mining

March 26, 2009

If we take seriously the results of more than a century of research on a possible genetic background of general intelligence (IQ), then a different distribution of allele frequencies in different populations and social strata should be expected (see "Major genes of general intelligence":www.v-weiss.de/majgenes.html and the final chapter of "Gene frequencies underlying national IQ means":www.v-weiss.de/calibration.html).

The present state of knowledge allows a first attempt to search for a major gene locus of IQ by data mining. From the total of 76690 nonsynonymously coding SNPs in the HapMap database, I have used the database SNPlogic to filter out 204 SNPs fitting within the expected ranges of allele frequencies in samples of European (CEU), Chinese (CHB), Japanese (JPT) and Sub-Saharan Yoruba (YRI) populations. Because among Europeans the frequency of the minor allele underlying high IQ in the homozygous state should not exceed 0.30, homozygosity by pure chance (0.30 x 0.30) can be expected in less than 0.10 of cases. Assuming (without IQ testing) that Craig Venter is a proband of high IQ, I used the published “Craig Venter genome”:http://jimwatsonsequence.cshl.edu/cgi-perl/gbrowse/cvsequence/ by looking for homozygosity of the rare allele and reduced in this way the list of candidate SNPs from 204 to 22.

Theoretically, by adding a second high-IQ proband with decoded genome data of similar quality to that of Craig Venter’s, this list could be further reduced by a factor of 0.10 to about 2, including the sought-after major gene locus. However, the next step, to exclude from these 22 at least the SNPs where James Watson is homozygous for the opposite (common) allele, led to no result at all. For 10 candidate SNPs there are no data in the James Watson genome database. All other SNPs are given as heterozygous with differing probabilities. In other words: The published James Watson genome database is completely unreliable.

However, with the help of Steven Pinker I came a step further. The Personal Genome Project and 23andMe are using the 500 K Affymetrix chip for genotyping. From analysing the Pinker data it follows that a high IQ gene cannot be on this chip. By data mining (and without any funding) I am replicating in this way the completely negative results of Plomin et al., who found no replicable correlation between any SNP on the 500 K Affymetrix chip and IQ. Contrary to Plomin, who is of the opinion that there exists no single gene with any substantial contribution to IQ, I drew the conclusion from his research that he always searched with insufficient methods at the wrong places.

At present, after the exclusion of all the SNPs which are on the 500 K Affymetrix chip, there remain the following 11 or 12 SNPs as candidates for a major gene locus of IQ (listed below with the Venter genotype and HapMap population frequencies of this allele):

rs238234 CAMTA2 Venter CC CEU 0.17 CHB 0.50 JPT 0.53 YRI 0.03

rs428785 ADAMTS1 Venter CC CEU 0.22 CHB 0.57 JPT 0.59 YRI 0.03

rs1919128 C2orf16 Venter GG CEU 0.24 CHB 0.56 JPT 0.62 YRI 0.04

rs2584625 probably identical with rs2727288 FTSJ3 Venter GG and/or TT CEU 0.38 CHB 0.40 JPT 0.46 YRI 0.00

rs3095726 ZNF573 Venter TT CEU 0.18 CHB 0.58 JPT 0.72 YRI 0.00

rs6961834 tcag7956 Venter TT CEU 0.37 CHB 0.48 JPT 0.44 YRI 0.06

rs4883918 DIS3 Venter CC CEU 0.19 CHB 0.51 JPT 0.51 YRI 0.03

rs12507582 DDX60L Venter TT CEU 0.31 CHB 0.63 JPT 0.70 YRI 0.03

rs12764004 BMS1 Venter AA CEU 0.20 CHB 0.48 JPT 0.34 YRI 0.03

rs13151700 DDX60L Venter GG CEU 0.32 CHB 0.60 JPT 0.70 YRI 0.05

rs17078347 PCDH24 Venter AA CEU 0.21 CHB 0.39 JPT 0.38 YRI 0.00

For some of these 10 genes very little is known. For others, for example CAMTA2 (now on the Illumina 1M chip), the present state of knowledge suggests they may be involved in information processing. The protein encoded by this gene is a member of the calmodulin-binding transcription activator (CAMTA) protein family and may function as a transcription factor that responds to calcium signalling by directly binding to calmodulin.

One should be aware that the Asian samples drawn from Beijing and Tokyo are not socially representative in any way. The Chinese HapMap (CHB) sample comes from a Beijing university resident academic population. Because little or nothing is known on the social representativeness of such population samples in general, thresholds for filtering out candidate SNPs could only be set using assumptions that could be wrong. Other sources of possible error are the incompleteness and unreliability of current databases.

After surviving two attacks of non-Hodgkin lymphoma, I had to retire and no longer have access to laboratory facilities of my own. Therefore, I am calling on colleagues all over the world to reduce this list of 10 candidate high-IQ genes to one or none. If the sought-after gene should actually be among these 10, we should all be surprised. I myself would rather expect a deletion and copy number variation (CNV) should or could underlie major IQ differences and not a single nucleotide polymorphism. However, at present, databases do still not contain population frequencies of CNV.

It should be demonstrated that by applying the powerful logic of genetics and available knowledge it is already possible to put forward reasonable hypotheses on IQ genes simply by data mining. Within a few years we will have open access to more and better databases. Therefore, we should be confident that a true breakthrough in IQ genetics is imminent, even without any funding and despite all political opposition and repression of this type of research. After their retirement, old men have nothing to fear anymore.

Without the help of Ivan Smirnov (SNPlogic), Steven Pinker and Andrew Walsh this data mining would not have been possible.

The investigation of pooled DNA samples by Plomin et al. has led to nothing at all. His opinion that high IQ is caused by a very large number of genes each accounting for a very small percentage of variance is nothing else than the declaration of complete bankruptcy of his approach. It is practically equivalent with the politically motivated point of view that there are no genes at all underlying general intelligence. I see no decisive difference.

Of course, array-based homozygosity mapping of candidate genes regions and high-resolution microarray-based comparative genomic hybridization (array CGH) within consangineous families, in order to detect such regions, needs fare more money and effort than investigating pooled DNA.