Using statistical and knowledge-based approaches for literature-based discovery

The explosive growth in biomedical literature has made it difficult for researchers to keep up with advancements, even in their own narrow specializations. While researchers formulate new hypotheses to test, it is very important for them to identify connections to their work from other parts of the literature. However, the current volume of information has become a great barrier for this task and new automated tools are needed to help researchers identify new knowledge that bridges gaps across distinct sections of the literature. In this paper, we present a literature-based discovery system called LitLinker that incorporates knowledge-based methodologies with a statistical method to mine the biomedical literature for new, potentially causal connections between biomedical terms. We demonstrate LitLinker’s ability to capture novel and interesting connections between diseases and chemicals, drugs, genes, or molecular sequences from the published biomedical literature. We also evaluate LitLinker’s performance by using the information retrieval metrics of precision and recall.     

Analysing gene expression data from DNA microarrays to identify candidate genes

Microarray data analysis can be divided into two tasks: grouping of genes to discover broad patterns of biological behaviour, and filtering of genes to identify specific genes of interest. Whereas the gene-grouping task is largely addressed by cluster analysis, the gene-filtering task relies primarily on hypothesis testing. This review article surveys analytical methods for the gene-filtering task. Various types of data analysis are discussed for four basic types of experimental protocols: a comparison of two biological samples; a comparison of two biological conditions; each represented by a set of replicate samples; a comparison of multiple biological conditions; and analysis of covariate information.     

Genomic convergence to identify candidate genes for Alzheimer Disease on chromosome 10

A broad region of chromosome 10 (chr10) has engendered continued interest in the etiology of late‐onset Alzheimer Disease (LOAD) from both linkage and candidate gene studies. However, there is a very extensive heterogeneity on chr10. We converged linkage analysis and gene expression data using the concept of genomic convergence that suggests that genes showing positive results across multiple different data types are more likely to be involved in AD. We identified and examined 28 genes on chr10 for association with AD in a Caucasian case‐control dataset of 506 cases and 558 controls with substantial clinical information. The cases were all LOAD (minimum age at onset ≥60 years). Both single marker and haplotypic associations were tested in the overall dataset and 8 subsets defined by age, gender, ApoE and clinical status. PTPLA showed allelic, genotypic and haplotypic association in the overall dataset. SORCS1 was significant in the overall data sets (p=0.0025) and most significant in the female subset (allelic association p=0.00002, a 3‐locus haplotype had p=0.0005). Odds Ratio of SORCS1 in the female subset was 1.7 (p<0.0001). SORCS1 is an interesting candidate gene involved in the Aβ pathway. Therefore, genetic variations in PTPLA and SORCS1 may be associated and have modest effect to the risk of AD by affecting Aβ pathway. The replication of the effect of these genes in different study populations and search for susceptible variants and functional studies of these genes are necessary to get a better understanding of the roles of the genes in Alzheimer disease. 30, 463–471, 2009. © 2008 Wiley‐Liss, Inc.     

A new evaluation methodology for literature-based discovery systems

While medical researchers formulate new hypotheses to test, they need to identify connections to their work from other parts of the medical literature. However, the current volume of information has become a great barrier for this task. Recently, many literature-based discovery (LBD) systems have been developed to help researchers identify new knowledge that bridges gaps across distinct sections of the medical literature. Each LBD system uses different methods for mining the connections from text and ranking the identified connections, but none of the currently available LBD evaluation approaches can be used to compare the effectiveness of these methods. In this paper, we present an evaluation methodology for LBD systems that allows comparisons across different systems. We demonstrate the abilities of our evaluation methodology by using it to compare the performance of different correlation-mining and ranking approaches used by existing LBD systems. This evaluation methodology should help other researchers compare approaches, make informed algorithm choices, and ultimately help to improve the performance of LBD systems overall.     

Nested Patch PCR enables highly multiplexed mutation discovery in candidate genes

Medical resequencing of candidate genes in individual patient samples is becoming increasingly important in the clinic and in clinical research. Medical resequencing requires the amplification and sequencing of many candidate genes in many patient samples. Here we introduce Nested Patch PCR, a novel method for highly multiplexed PCR that is very specific, can sensitively detect SNPs and mutations, and is easy to implement. This is the first method that couples multiplex PCR with sample-specific DNA barcodes and next-generation sequencing to enable highly multiplex mutation discovery in candidate genes for multiple samples in parallel. In our pilot study, we amplified exons from colon cancer and matched normal human genomic DNA. From each sample, we successfully amplified 96% (90 of 94) targeted exons from across the genome, totaling 21.6 kbp of sequence. Ninety percent of all sequencing reads were from targeted exons, demonstrating that Nested Patch PCR is highly specific. We found that the abundance of reads per exon was reproducible across samples. We reliably detected germline SNPs and discovered a colon tumor specific nonsense mutation in APC, a gene causally implicated in colorectal cancer. With Nested Patch PCR, candidate gene mutation discovery across multiple individual patient samples can now utilize the power of second-generation sequencing.     

A functional candidate screen for coeliac disease genes

It is increasingly evident that different inflammatory disorders show some overlap in their pathological features, concurrence in families and individuals, and shared genetic factors. This might also be true for coeliac disease, a chronic inflammatory disorder of the gastrointestinal system, which shares two linkage regions with inflammatory bowel disease: on chromosome 5q31 (CELIAC2 and IBD5) and 19p13 (CELIAC4 and IBD6). We hypothesised that these regions contain genes that contribute to susceptibility to both disorders. The overlapping 5q31 region contains only five positional candidate genes, whereas the overlapping 19p13 region has 141 genes. As the common disease gene probably plays a role in inflammation, we selected five functional candidate genes from the 19p13 region. We studied these 10 positional and functional candidate genes in our Dutch coeliac disease cohort using 44 haplotype tagging single-nucleotide polymorphisms. Two genes from 19p13 showed a small effect on familial clustering: the cytochrome P450 F3 gene CYP4F3 (P(nominal) 0.0375, odds ratio (OR) 1.77) and CYP4F2 (P(nominal) 0.013, OR 1.33). CYP4F3 and CYP4F2 catalyse the inactivation of leukotriene B4 (LTB4), a potent mediator of inflammation responsible for recruitment and activation of neutrophils. The genetic association of LTB4-regulating gene variants connects the innate immune response of neutrophil mobilisation with that of the established Th1 adaptive immunity present in coeliac disease patients. The findings in coeliac disease need to be replicated. Expanding genetic association studies of these cytochrome genes to other inflammatory conditions should reveal whether their causative influence extends beyond coeliac disease.     

Rare copy number variation discovery and cross-disorder comparisons identify risk genes for ADHD

Attention deficit hyperactivity disorder (ADHD) is a common and persistent condition characterized by developmentally atypical and impairing inattention, hyperactivity, and impulsiveness. We identified de novo and rare copy number variations (CNVs) in 248 unrelated ADHD patients using million-feature genotyping arrays. We found de novo CNVs in 3 of 173 (1.7%) ADHD patients for whom we had DNA from both parents. These CNVs affected brain-expressed genes: DCLK2, SORCS1, SORCS3, and MACROD2. We also detected rare inherited CNVs in 19 of 248 (7.7%) ADHD probands, which were absent in 2357 controls and which either overlapped previously implicated ADHD loci (for example, DRD5 and 15q13 microduplication) or identified new candidate susceptibility genes (ASTN2, CPLX2, ZBBX, and PTPRN2). Among these de novo and rare inherited CNVs, there were also examples of genes (ASTN2, GABRG1, and CNTN5) previously implicated by rare CNVs in other neurodevelopmental conditions including autism spectrum disorder (ASD). To further explore the overlap of risks in ADHD and ASD, we used the same microarrays to test for rare CNVs in an independent, newly collected cohort of 349 unrelated individuals with a primary diagnosis of ASD. Deletions of the neuronal ASTN2 and the ASTN2-intronic TRIM32 genes yielded the strongest association with ADHD and ASD, but numerous other shared candidate genes (such as CHCHD3, MACROD2, and the 16p11.2 region) were also revealed. Our results provide support for a role for rare CNVs in ADHD risk and reinforce evidence for the existence of common underlying susceptibility genes for ADHD, ASD, and other neuropsychiatric disorders.     

Using microarray analysis to identify genes and pathways that regulate fetal hemoglobin levels

Increased levels of fetal hemoglobin (HbF: α2γ2) can ameliorate the clinical severity of the β-hemoglobinopathies. Microarray analysis represents a powerful approach to identify novel genetic factors regulating the γ-globin gene. Gene expression profiling was previously performed on 14 individuals with high or normal HbF levels to identify the genetic factors that control γ-globin gene expression. To obtain more accurate and reliable results, our results were combined with public microarray dataset GSE22109 deposited in the Gene Expression Omnibus database. Annotation of case versus control samples was taken directly from the microarray documentation. The differentially expressed genes (DEGs) were obtained and were deeply analyzed by bioinformatics methods. Combined with our own chip expression data, potential genes HBE1, TFRC, and CSF2 were selected out for subsequent qRT-PCR validation. A total of 184 DEGs were identified from GSE22109 and the protein–protein interaction network was constructed. Gene set enrichment analysis showed that the hematopoietic cell lineage pathway overlaps in the two datasets. HBE1, CSF2, and TFRC were confirmed by qRT-PCR. Our results suggest novel candidate genes and pathways associated with the γ-globin gene expression.     

Linkage analysis of candidate regions for coeliac disease genes

A strong HLA association is seen in coeliac disease [specifically to the DQ(alpha1*0501,beta1*0201 heterodimer], but this cannot entirely account for the increased risk seen in relatives of affected cases. One or more genes at HLA-unlinked loci also predispose to coeliac disease and are probably stronger determinants of disease susceptibility than HLA. A recent study has proposed a number of candidate regions on chromosomes 6p23 (distinct from HLA), 6p12, 3q27, 5q33.3, 7q31.3, 11p11, 15q26, 19p13.3, 19q13.1, 19q13.4 and 22cen for the location of a non-HLA linked susceptibility gene. We have examined these regions in 28 coeliac disease families by linkage analysis. There was excess sharing of chromosome 6p markers, but no support for a predisposition locus telomeric to HLA. No significant evidence in favour of linkage to coeliac disease was obtained for chromosomes 3q27, 5q33.3, 7q31.3, 11p11, 19p13.3, 19q13.1, 19q13.4 or 22cen. There was, however, excess sharing close to D15S642. The maximum non-parametric linkage score was 1.99 (P = 0.03). Although the evidence for linkage of coeliac disease to chromosome 15q26 is not strong, the well established association between coeliac disease and insulin dependent diabetes mellitus, together with the mapping of an IDDM susceptibility locus (IDDM3) to chromosome 15q26, provide indirect support for this as a candidate locus conferring susceptibility to coeliac disease in some families.      

Functional genomics in rat models of hypertension: using differential expression and congenic strains to identify and evaluate candidate genes

Hypertension is a leading contributor to cardiovascular diseases such as heart attack and stroke. Genetic and environmental factors contribute to the development of hypertension. Animal models have been developed to study the genetic contributions to blood pressure (BP) regulation and to identify chromosomal regions harboring candidate genes causative of differences in BP regulation (i.e., BP quantitative trait loci [QTL]). Advances in both mammalian genome projects and global gene expression analysis present opportunities to study functional genomics in these animal models. In this article, novel approaches for designing experiments and interpreting global gene expression data using the Dahl salt-sensitive hypertension rat model are presented. We describe two-step screening protocols that can be used to identify BP QTL candidate genes. Genetically determined expression differences are identified in the target organs of inbred strains of contrasting phenotype in the first screen. Expression patterns in a panel of congenic strains or expression differences stemming from gene x environment interactions are examined in the second screen. Chromosomal locations of these genes can then be examined to determine whether they map to BP QTL-containing regions. Another approach is to study the expression of genes identified from public databases to be located within BP QTL-containing congenic regions. Several candidate genes have been identified using these strategies.     

Mutations and polymorphisms of Hirschsprung disease candidate genes in Thai patients

Mutation and polymorphism data for Hirschsprung disease (HSCR) varies among ethnic groups. Single nucleotide polymorphisms (SNP) of RET proto-oncogene (RET) were recently shown to be associated with the disease, and with disease severity, in different populations. In this study, comprehensive analysis of RET, GDNF, EDNRB, ET-3, and SOX-10 genes among sporadic HSCR in Thailand was conducted by standard PCR-SSCP, RFLP, and sequencing methods. Of 41 patients, 30 cases had rectosigmoid disease (RSD) and 11 cases were assigned to the long-segment disease (LSD) group. Four missense mutations of RET, S100M, R231H, T278N, and G533S, were identified in three patients. One novel missense mutation, V111Q, was detected in EDNRB. For ET-3, two novel missense mutations, D166E and C173R, occurred concomitantly in a patient. The incidence of missense mutation was significantly higher in our female HSCR patient than in the male counterpart. Statistical analysis of the SNPs revealed a significant difference between allele distribution of RET L769L in patients in the LSD and RSD groups. The predominant genotype construct of RET A45A/L769L in our HSCR was GG/GG, which is obviously different from results from all previous studies. The GG/GG genotype construct was associated with RSD and with males. The study also detected a variant allele of RET S836S which has never been reported in Asian cohorts.     

Endometriosis: candidate genes

Endometriosis is generally regarded as a benign disease but it does exhibit some characteristics reminiscent of malignancy. This raises the possibility that, like malignant diseases, the development of endometriosis may involve the acquisition of somatic genetic alterations in genes that regulate cell growth and differentiation. Studies over the past few years have substantiated this view with the identification of a variety of genetic abnormalities usually only associated with malignancies. Our own studies have shown that genetic alterations, as shown by loss of heterozygosity, are relatively common in endometriosis implying that tumour suppressor gene inactivation is likely to be involved in the proliferation and maintenance of all endometriotic implants. We have also shown by DNA fingerprinting that endometriotic lesions found adjacent to ovarian cancers have a common lineage, reinforcing the compelling histological and epidemiological data that endometriosis is a precursor of endometrioid and clear cell ovarian cancers. It is now well accepted that susceptibility to endometriosis may also involve an inherited genetic component. Studies aimed at identifying the predisposing genes are still in their infancy but should eventually provide invaluable insights into the pathology and aetiology of endometriosis.     

Genetic studies to identify genes underlying menopausal age

Menopausal age is important as a retrospective marker for ovarian senescence, an early menopausal age is associated with an increased risk of cardiovascular diseases and osteoporosis, whereas a later menopausal age has been associated with an increased risk of breast cancer. The worldwide average for age at natural menopause is approximately 51 years and is more or less normally distributed with a range roughly between 40 and 60 years. Environmental factors explain only a small part of the variance and it has been proposed that genetic factors are the main source of variation. Menopausal age may be considered a continuous complex trait. Complex traits are defined as traits that are influenced by both multiple genetic and environmental factors. A category of complex traits comprises those that are measured on a continuous scale. The genomic loci that make up the genetic component are called 'quantitative trait loci' or QTLs. The first linkage study on menopausal age suggests that the involvement of the X-chromosome may not be limited to premature ovarian failure (POF), but may influence the broader spectrum of menopausal age. A potentially new locus for variation in menopausal age was allocated to chromosome 9. Further studies need to identify new candidate genes to help unravel the pathophysiology of menopausal age. It is becoming increasingly clear that, in any speciality, it should be acknowledged that genetic factors are involved in many traits and that uncovering these factors may provide insight into pathogenesis and ultimately advance prevention and treatment of disease. In this review we discuss methods and basic principles of gene finding for such traits, exemplified by menopausal age as phenotype. Furthermore, we give an overview of the state of the art of candidate gene studies and linkage studies.     

Prioritizing candidate disease genes by network-based boosting of genome-wide association data

Network "guilt by association" (GBA) is a proven approach for identifying novel disease genes based on the observation that similar mutational phenotypes arise from functionally related genes. In principle, this approach could account even for nonadditive genetic interactions, which underlie the synergistic combinations of mutations often linked to complex diseases. Here, we analyze a large-scale, human gene functional interaction network (dubbed HumanNet). We show that candidate disease genes can be effectively identified by GBA in cross-validated tests using label propagation algorithms related to Google's PageRank. However, GBA has been shown to work poorly in genome-wide association studies (GWAS), where many genes are somewhat implicated, but few are known with very high certainty. Here, we resolve this by explicitly modeling the uncertainty of the associations and incorporating the uncertainty for the seed set into the GBA framework. We observe a significant boost in the power to detect validated candidate genes for Crohn's disease and type 2 diabetes by comparing our predictions to results from follow-up meta-analyses, with incorporation of the network serving to highlight the JAK-STAT pathway and associated adaptors GRB2/SHC1 in Crohn's disease and BACH2 in type 2 diabetes. Consideration of the network during GWAS thus conveys some of the benefits of enrolling more participants in the GWAS study. More generally, we demonstrate that a functional network of human genes provides a valuable statistical framework for prioritizing candidate disease genes, both for candidate gene-based and GWAS-based studies.     

Human cochlear expressed sequence tags provide insight into cochlear gene expression and identify candidate genes for deafness

To identify candidate genes for human hearing disorders and to understand better human hearing at the molecular level, we constructed a human cochlear cDNA library. An aliquot of the unsubtracted cochlear library was contributed to the IMAGE Consortium at Lawrence Livermore National Laboratory for the generation of expressed sequence tags (ESTs) by the Merck/WashU EST project. Over 4000 ESTs were developed from the cochlear cDNA library and deposited in the GenBank EST database. Sequence clustering shows that the majority of clones are in low copy numbers, demonstrating the high complexity of the library. The sequences of 1388 cochlear ESTs (33%) match 517 known human genes. Among these are genes previously shown to cause both syndromic and non- syndromic hearing loss. A number of the cochlear ESTs show high homology to non-human genes, suggesting new gene family members or human homologs of animal genes. We also report the chromosomal map positions of 437 cochlear ESTs. These provide positional candidate genes for 18 different non-syndromic hearing disorders. A Human Cochlear EST Database web site (http://www.bwh.partners. org/pathology ) has been created to provide access to the cochlear clone data for gene discovery investigations.