Analyzing the most frequent disease loci in targeted patient categories optimizes disease gene identification and test accuracy worldwide

The most frequent listed disease locations comprise the largest proportion of testable disease causing mutations worldwide. Estimated disease frequencies found in all tested categories were derived from our calculations based upon reported patient and general population data [Additional file 1: Table S1, Additional file 2: Table S2 and Additional file 3: Table S3]. The most frequent disease alleles for cystic fibrosis and the hemoglobinopathies are reported to have been selected by heterozygous advantage [26‐28]. In contrast, the other most frequent autosomal recessive genetic disease genes have many unique alleles but none were reported to have sufficiently frequent mutations that comprise a major proportion of all mutations (Additional file 1: Table S1A; Ref. [29]). Thus determining whether a variant gene sequence is normal at a frequent disease gene site should include a comparison to all confirmed mutations.

Other frequent diseases in ethnic populations result from a limited founder pool with offspring who regularly select a partner from among the offspring of all the founders [Additional file 1: Table S1F]. These ethnic populations can be tested effectively by targeting the few most frequent mutations in the founders’ rare disease alleles. Initially a population disease frequency may be overestimated when sampled from a region with a higher carrier frequency. Disease frequencies in founder populations can also skew panethnic population frequencies when offspring migrate together to specific geographic regions like in the United States where more centrally located populations can more often trace their ancestors to citizens who arrived earlier in the nation’s history.

The most extensive available disease frequency estimates with the broadest geographical distributions were found on the GeneTests website in the GeneReviews chapters. This study considered all the >2500 listed disease genes [1] for inclusion in at-risk and affected patient test categories. The disease list selected was further organized by transmission category: (A) autosomal recessive, (B) X-linked, (C) autosomal dominant, (D) Y-linked, (E) mitochondrial, and (F) frequent diseases in specific populations [Additional file 1: Table S1A-F, Tables 1 and 4]. From the 534 listed and updated GeneTests frequencies, 125 diseases were selected that are each reported to affect at least 1 in 100,000 Caucasian individuals [Additional file 1: Table S1A,B,C,D] including 122 diseases reported to affect at least 1 in 100,000 people worldwide [Additional file 1: Table S1A,B,C,D, excluding #1, 2, 38] and 55 disease genes frequent in specific populations [Additional file 1: Table S1F].

Additional estimates of disease gene frequencies have been derived from our tested at-risk and abnormal patient samples [Additional file 2: Table S2, Col 2.3.4; Additional file 3: Table S3]. Initially the abnormal karyotype categories are listed according to decreasing frequency in clinically reported products of conception [Additional file 2: Table S2A]. These karyotype results were then reorganized according to estimated severity to facilitate comparison of the remaining viable karyotypic abnormalities as gestation progresses [Additional file 2: Table S2B]. Calculated general population disease frequencies include 50% of abnormal karyotyped products of conception [POC] and 0.64% of abnormal karyotypes in newborns [Additional file 2: Table S2B, Col 2,5, Bottom]. In contrast, for prenatally sampled fetuses tested by chorionic villus sampling [CVS], abnormal karyotypes were reported in 3.1% completed by 1992 in San Francisco [21] compared to 33% completed more recently in Ohio (Additional file 2: Table S2B, Col 3]). Note also the increased number of abnormal chromosome categories in the more recent sampled amniocenteses [Additional file 2: Table S2B, Col 4, Right, underlined).

Then the most frequent 40 aneuploid locations spanning >400,000 basepairs characterized by our microarrays were organized according to the frequencies of each abnormality among our 1265 tested patients [Additional file 3: Table S3A]. These frequencies among Texas’ patients (T) were subsequently reorganized according to chromosome location within 3 categories to facilitate comparison of (1) the 13 confirmed gene loci with estimated general population frequencies of at least 1 in 100,000, (Additional file 3: Table S3B, Top, Right Column, Ref. [1]) (2) the additional reported chromosome regions that result in altered neurocognitive development with or without other abnormalities ((K); Additional file 3: Table S3B, Col 4; Ref. [24]), and (3) the remaining 12 abnormal clinically reported loci. Nearly all frequent deletions listed at the top of Additional file 3: Table S3B also result in altered developmental delay. When testing for all these abnormalities, 10% of patients were reported positive for gene deletion or duplication in this Texas’ cohort.

The frequencies of six disease categories with available general population frequencies [Additional file 1: Table S1A,B,C; Additional file 2: Table S2, Newborn; Additional file 3: Table S3B,Top] were calculated and graphed for ready interpretation [Figure 1A,B,C]. Note that the largest proportion of reported disease frequencies is found in the first subcategory [1/1 to 1/25,000]. These proportions diminish rapidly as the affected patient frequency decreases in increments to [<1/75,000 to 1/100,000]. This consistent trend can be applied to optimally select disease inclusion frequency as ongoing test experience accumulates. These graphs demonstrate that any platform that tests the most frequent disease genes according to physician selected or peer reviewed disease categories will expedite reporting the largest proportion of clearly positive and negative test results.

Altogether the first 227 [125 + 54 + 48; Additional file 1: Table S1, Col.1; Additional file 2: Table S2, Table 5; Additional file 3: Table S3A] most frequent Caucasian diseases at 263 loci [227 + 36; See Additional file 1: Table S1A,B,C,D, Column 4] result from mutations involving ~1.2% of the ~22,000 human genes [263/22,000 = 1.19%]. This total is 224 diseases for worldwide patients [122 + 54 + 48]. These values facilitated deriving the frequencies of at-risk asymptomatic couples carrying autosomal recessive [Table 4.A.2] and X-linked recessive diseases (Table 4.B.2) along with the dominant and Y-linked patient categories [Tables 1 and 4].

Frequent abnormal alleles in populations that are likely to select a partner from within their own ethnic group account for virtually all of the remaining homozygous autosomal recessive affected conceptuses. Altogether GeneTests reported 55 diseases that each affects at least 1 in 100,000 patients in specific ethnic populations [Additional file 1: Table S1F]. For the general U.S. population, the first five listed ethnicities each comprise at least 2% of the total U.S. population and together have 31 ethnic specific diseases that can be added readily to the general population screening test [Additional file 1: Table S1F, #1-31]. Most ethnic gene mutations require substantially less effort to test because these typically include very few frequent abnormal alleles.

The summary of these individual calculated affected and carrier disease frequencies are listed [Table 4] as well as the composite rates for Caucasian and Worldwide patient categories [Table 1]. These most frequent disease lists were constructed to facilitate physician-selected disease testing that could be ordered for four patient categories: (1) patients selecting a partner or reproducing, (2) at-risk conceptuses, (3) abnormal newborns and minors, and (4) affected adults [Tables 1 and 2].

Currently next generation platforms and their modifications can be used immediately to rapidly sequence the total exome or ~4800 disease related genetic sites for pregnant couples and affected newborns and adults. Analyzed genes can be selected from among the total results. Fetuses at a 1 in 4 risk of a known gene defect are tested for this risk first. Karyotypes are being ordered initially in Ohio for small invasive prenatal samples, very late gestation fetuses and 90% of refrigerated POCs that are cultured and karyotyped. Other facilities order microarrays [31] for sufficient CVS and amniocyte samples and up to 40% of POC samples that can not be cultured.Platforms that complete multiple test categories as well as the most important fetal test will continue to be developed and selected from the most reliable source as test platforms evolve and updated databases are constructed and maintained (Refs. [1,32]; Peer-reviewed publications).

Akron Children’s product of conception [POC] protocol includes a pathologist’s examination, description, biopsy, histology, and submission of freshly biopsied chorionic villi for karyotyping. Chorionic villus biopsies are preferred to other fetal tissues because 90% of viable submitted fetal samples maintain cell viability when refrigerated up to 5 days prior to cell culture or DNA analysis compared to the poorer viability of other fetal tissues. Histologically prepared villi are analyzed for mole and partial mole morphology (Additional file 4: Table S4). The cytogenetics laboratory assures dissection of the villi to 90-95% purity for karyotyping and 98-100% purity for molecular analyses. After digestion and culturing, chorionic villus cells from 90% of submitted placental samples have been karyotyped. (Additional file 2: Table S2A). Following a preliminary report, polymorphic DNA identity testing confirms the fetal origin of the cultured samples prior to a final report. Of the remaining 10% of failed cultures, 7% are sufficiently intact to analyze interphase nuclei by FISH for chromosome 13, 16, 18, 21, 22, X, and Y aneuploidy comprising ~63% of all chromosome abnormalities in our karyotyped POCs. These 50% of cultures with abnormal reported karyotypes among all karyotyped POCs are listed in order of frequency for each category [Additional file 2: Table S2A, Col 2]. Given the 90% of samples karyotyped, these abnormal frequencies are interpreted to reflect typical POC population frequencies. ACOG recommends using microarrays for POC samples when 20%-50% of samples fail to grow in cuture and cannot be karyotyped [34]. Given 90% of our POCs are karyotyped according to our protocol, we have karyotyped all cultured POCs to identify tetraploidy from diploid genomes and balanced or complex categories of chromosome abnormalities currently missed by microarrays and genomewide sequencing.

Sampled at-risk fetuses have been karyotyped following CVS or amniocentesis to serve as the clinically standard test for microscopically visible chromosome abnormalities. For comparison we added the most recent 12 years of prenatal karyotypes in Ohio to our published 25 years of prenatal karyotypes in San Francisco from 1970 to 1995. This comparison found a 10.6-fold increase in more recent abnormal CVS karyotypes [3.1% to 33%] and a 3.2-fold increase in more recent abnormal amniocyte karyotypes (2.3% to 7.2%; Ref. [21]; Additional file 2: Table S2A, Col 3,4) These substantially higher abnormal frequencies in Akron’s more recent samples are consistent with those published recently by the American College of Obstetrics and Gynecology [34].

Subsequently these results were organized by severity to emphasize the evolution of surviving abnormal karyotypes as the time after conception increases [Additional file 2: Table S2B)]. These POC frequencies were compared to the reported frequencies in CVS, amniocentesis, and newborn karyotypes (Additional file 2: Table S2B, Col 2,3,4,5; Ref. [21]). Additional categories were observed in POC specimens: (1) triploid to aneuploid,(2) tetraploid to aneuploid, and (3) aneuploid to tetraploid [Additional file 2: Table S2B Col 2, Top, Underlined]. Additional categories were also observed in the more recent amniocyte samples over those previously published: (1) double aneuploidy, (2) isochromosomes, (3) complex abnormalities with two or more abnormal chromosome categories, (4) diploid to tetraploid, (5) trisomy 16 and 22, (6) monosomy 11 and 21, and (7) mosaic karyotypes [Additional file 2: Table S2B, Col 4, Underlined, Italicized].

The 54 selected sites reported to comprise the most frequent chromosome regions involved in abnormal chromosome copy number (Table 5, Reproduced from Ref. [30]; Derived from [32]) were found to have identified all the abnormal chromosome copy number regions identified in the listed abnormal POC, CVS, and amniocyte karyotypes in Akron [Additional file 2: Table S2B]. Because 2,500,000 site polymorphic microarrays only detect abnormal copy number, the frequency of abnormality detection would not have been increased by this assay, but the chromosome region spanned by the copy number change would have been delineated. These abnormal 54 aneuploid loci can be identified on any platform that (1) detects a sufficient number of the most informative adjacent single, di- and tetra-nucleotide polymorphisms and quantifies the relative and total number of times each of these sites were sequenced or (2) the relative number of targets at the aneuploid site compared to a normal diploid control region. Candidate platforms include rapid sequencers and microarrays. Following initial identification of an aneuploid chromosomal gene region on a 54 site test, the extent of the aneuploid gene region could be delineated readily by completing a typical polymorphic genomewide microarray or a ~4800 disease gene sequence.

Alternative platforms and karyotyping should be compared carefully for different applications as tests evolve. For instance, among the ~50% of karyotypically abnormal POC samples, ~45% of these abnormal samples would have been detected by either a 54 site or a 2,500,000 site genomewide test platform provided the control DNA includes an intact Y chromosome [Additional file 2: Table S2B, not italicized]. The same platforms would have defined copy number changes in 5.7% of the 7.2% [(79%); (7.2%-1.54%)] abnormal amniocyte karyotypes and in ~25.7% of the 33% [(78%); (33%-7.3%)] current abnormal CVS karyotypes [Additional file 2: Table S2B].

Compare this to microarrays with 2,500,000 sites that precisely map unknown and previously detected unbalanced rearrangements and submicroscopic aneuploidy. These extensive microarrays are particularly useful in delineating the 6% of inherited cases with undetected submicroscopic deletions or duplications not detected in karyotypes of a conceptus with an apparently balanced chromosomal rearrangement inherited from a normal carrier parent [35]. In spite of the propensity of unequal chromosome recombinations among the potentially confounding ~2,370,000 copy number variants [3,36], balanced rearrangements are beginning to be detected by improved genomewide platforms and sophisticated computer analysis [17,37]. Karyotyping is the only means to detect diploid to tetraploid mosaicism and tetraploidy arising from diploid cells. As microarray, sequencing, and flow sorting methods for detecting balanced abnormalities are being validated in additional patients with substantial analytic resources,[17] standard karyotyping is still the preferred method for detecting balanced and complex rearrangements in a few viable cells.

Selecting the most optimal platform for any individual fetus will depend upon available ongoing test deveopment and available protocols through referral laboratories. At our location we karyotype all CVS samples because of the 33% found to have abnormal karyotypes. Currently we karyotype amniocytes for substantially abnormal ultrasounds and select the fetal samples to be sent for microarray analysis following following rapid FISH overnight. Our first sample designated for microarray testing was positive for trisomy 21 by Rapid FISH and subsequent karyotyping.

Pediatricians following standard of care can decide to test symptomatic newborns and children for all frequent genetic disease categories that may define the abnormal phenotype. These disease categories include the autosomal dominant diseases resulting in ~1 in 123 affected newborns [Additional file 1: Table S1C, Table 4C] and the frequent deletions and duplications causing an estimated ~1 in 549 affected newborns [1/1097 X 2; Additional file 3: Table S3A,B, Table 4E]. Together these and all previously mentioned genetic disease categories [Table 1.3] result in ~1 in 54 [1.86%] affected newborns worldwide and 1 in 52 [1.91%] Caucasian newborns.

Calculated dominant disease prevalence reflects a disease frequency that includes many heterozygous abnormal genotypes [2pq] and very few homozygous affected patients [q2] in randomly mating populations with reproducing patients. Mutations involving 60 of these frequent autosomal dominant diseases [Additional file 1: Table S1C, Table 4C] include a substantial proportion that resulted from de novo mutations [2% to 95%]. As anticipated, these de novo autosomal dominant mutations are often more severe than autosomal dominant diseases inherited from reproducing adults. If an autosomal dominant disease gene were identified in an affected minor, parents could then be counseled and offered testing to determine whether the minor’s disease gene is de novo [including parental germ line and somatic mosaicism], or resulted from substantial anticipation, variable penetrance, or variable expressivity. Testing these autosomal dominant diseases requires a platform with the sensitivity to detect single nucleotide mutations.

The submicroscopic deletion and duplication category was addressed by analyzing the prior 5 years of abnormal microarray results that each spanned >400,000 basepairs in our Texas’ patient cohort. These samples were submitted in order of highest to lowest frequency by Neonatologists, Pediatric Neurologists, Geneticists, Developmentalists, and Pediatricians. The 9.6% abnormal results [121/1265] were organized into three categories [Additional file 3: Table S3B]: (1) 13 loci with frequent deletions with reported normal and abnormal population frequencies that nearly all involved altered neurocognitive development [Additional file 3: Table S3B, Top], (2) at least 23 additional reported loci related to altered neurocognitive development with reported frequencies in abnormal patients [Additional file 3: Table S3B, Middle], and (3) 12 additional clinically significant loci including three chromosomal abnormalities. [Additional file 3: Table S3B, Bottom] About 1/2 of the 123 total deletions and duplications spanned 9 of the first 13 recurrent deletion loci. The other ~1/2 of the cases in categories 2 and 3 included 8 recurrent schizophrenia loci [24,38].

A 52-fold enrichment was found between the 1 in 21 affected patients selected for microarray testing in Texas and an estimated 1 in 1097 patients predicted in the general population by estimated disease frequencies [Additional file 3: Table S3B, Top]. This ~50-fold enriched frequency of most frequent deletions found among all patients submitted for microarray analysis illustrates the principle that testing clinically suspicious phenotypes substantially enhances the affected patient frequency among tested samples [Additional file 2: Table S2 and Additional file 3: Table S3]. Although a chromosome abnormality was not suspected, 15 cases with trisomy 21, trisomy 18, or iso(12p) were identified by microarrays in our Texas cohort. Given the difficulty in identifying the exact genetic abnormality by the patient phenotype alone, selected karyotyping may be prudent.

Symptomatic adults can be tested for all disease categories mentioned previously as well as late onset genetic diseases following appropriate counseling and informed consent. These diseases include Parkinson’s disease, Alzheimer disease, and amyotrophic lateral sclerosis in worldwide populations and Huntington disease in Caucasians.

The 52 genetic diseases reported to be most frequent in specific populations listed in GeneTests have been incorporated to expedite optimal disease testing in regional laboratories (Additional file 1: Table S1F). Israeli laboratories will want to incorporate platforms with the most frequent Jewish mutations where 6,000,000 Israeli Jews reside, while laboratories in Sweden and Norway would include the Nordic mutations, labs in Quebec the French Canadian mutations, and Asian labs their regionally reported mutations. Worldwide, subpopulations of citizens within specific populations that typically select partners from their own ethnic groups would readily be served by incorporating the few frequent mutations for these diseases into the population wide screening test. These include Asian, Black, and Jewish United States citizens that each comprise at least 2% of the entire pan ethnic United States population (Additional file 1: Table S1F Group 3; Ref. [39]) and regional laboratories in central Pennsylvania, northeast Ohio, and Colorado where 200,000 Amish and Mennonite regional residents [0.08% of the total panethnic population] carry >50 specific gene mutations reported in these descendents of a few score of founders.

Currently interpreting genomewide exome sequencing of all ~22,000 genes requires several times the cost of sequencing. Dr. Hruban reported “a human exome has on average almost 36,000 variants, 45 percent of which are not in an SNP database and about 100 of which can cause loss of function” [40]. He concluded, “The potential power of next-generation sequencing for clinical testing is substantial. It will be a while before it is brought fully to the clinic…” This agrees with Chun et al. who identified ~90 potential disease causing sequences identified by exome analysis of all ~22,000 genes in three individuals [5]. When testing 100,000 patients for the frequent autosomal recessive cystic fibrosis mutations with 99.9% test accuracy per gene, a positive carrier test would include 3445 correct answers and 103 incorrect answers (Table 3, Top; Ref. [20]). In contrast, for the rare fumarase deficiency gene locus with the same test accuracy a positive carrier test would not only include 26 correctly analyzed carriers but also 100 incorrect false positive carrier test results [Table 3, Bottom]. Maintaining the principle of testing the most frequent diseases listed from most to less frequent in each selected category provides an objective basis to select the lowest frequency of any tested disease in each category.