Bioinformatics & biostatistics tools for monogenic and multifactorial disease investigation in consanguineous populations
2 likes578 views
The document discusses bioinformatics and biostatistics applications for examining monogenic and multifactorial diseases in consanguineous populations, highlighting the significance of inbreeding and genetic reference panels. It details methods for assessing inbreeding coefficients and the implications for identifying both rare monogenic and common multifactorial diseases. Specific case studies illustrate the power of homozygosity mapping and genomic tools in elucidating disease-related genetic factors within these populations.
![ Pedigree (fg) provides a value averaged over all possible lines of
descent: E(f) = fg
Large variability of the genome-based IBD around fg
1C: 95% interval = [0.02; 0.12] vs. fg=0.0625
fg=0.125 fg=0.0625 fg=0.0156 fg=0.0039
f
Pedigree- vs Genome-based IBD](https://image.slidesharecdn.com/lesoutilsbioinformatiquesetbiostatistiquespourltudedespopulationsconsanguines-161103165052/85/Bioinformatics-biostatistics-tools-for-monogenic-and-multifactorial-disease-investigation-in-consanguineous-populations-9-320.jpg)
![Genomically controlled homozygosity mapping
Leutenegger et al, AJHG, 2006
• H1: disease locus at position k
H0: position k is a random point of the individual’s genome
• Y set of marker genotypes → observed
X set of marker IBD statuses (0 or 1) → unobserved
For each affected individual: estimation of f, computation of the
posterior IBD probabilities at each marker k [P(Xk|Y)]
q: disease allele frequency; genetic model: fully penetrant recessive disease
• For a sample of independent affected inbred individuals:
FLOD = Σi FLODi
ffqf
YXqPYXP
FLOD q
kk
i
1
log
)1(
)0()1(
log 0→≈
−+
=+=
=
).(
).(
log
0
1
HdataobsP
HdataobsP
=Homozygosity mapping LOD score = HMLOD](https://image.slidesharecdn.com/lesoutilsbioinformatiquesetbiostatistiquespourltudedespopulationsconsanguines-161103165052/85/Bioinformatics-biostatistics-tools-for-monogenic-and-multifactorial-disease-investigation-in-consanguineous-populations-28-320.jpg)
Bioinformatics & biostatistics tools for monogenic and multifactorial disease investigation in consanguineous populations
- 1. Bioinformatics & biostatistics tools for monogenic and multifactorial disease investigation in consanguineous populations Mourad SAHBATOU Fondation Jean DAUSSET - CEPH, Paris mourad.sahbatou@cephb.fr Cours international Pasteur - Tunis – Octobre 2016
- 2. Outline 1. Consanguineous populations 1. Definition and characteristics 2. Measures of consanguinity 3. Consanguinity and Human Genetic Reference Panels 2. Human disease investigation in consanguineous populations 1. Rare monogenic diseases 2. Common multifactorial diseases
- 3. Consanguineous populations Populations where marriages between relatives are frequent • Populations where marriages between close relatives are encouraged for socio-economic or cultural reasons Saudi Arabia: 36% first cousins (Khlat, 1996) India-Karnataka: 21% uncle-niece (Bittles et al, 1992) • Populations isolated for cultural or geographical reasons where inbreeding results from a small number of founding individuals and low migration rates Cultural or religious isolates: Amish, Hutterites, Ashkenazi Jews, Gipsy Geographic isolates: Quebec, Iceland, Finland, Sardinia
- 4. Typical pedigree structures Population isolates Iceland Populations where marriages between close relatives are favored Saudi Arabia
- 5. Genetic consequences Marriages between relatives: → an individual may receive twice the same allele from an ancestor → identity by descent (IBD) Inbreeding coefficient: probability that 2 alleles at a random locus of the individual genome are IBD First cousin offspring, inbreeding from pedigree fg = (1/2)6 x 4 = 1/16 ≈ 6% Two alleles IBD
- 6. Within and between individual IBD • In population isolates »several links between two individuals • Identity by descent »of the two alleles in one individual : IBD within individual, inbreeding coefficient »of the alleles of two different individuals: IBD between individuals, kinship coefficient
- 7. Measuring IBD: pedigree or genome ? • Pedigree information is often incomplete, cannot go back more than a few generations »In consanguineous populations, pedigrees (if known) are large and complex • In recent years, production of large amounts of information from the genome (millions of SNPs) interest in estimating inbreeding and kinship from the genome, as the proportion of genome IBD
- 8. Distribution of IBD over the genome IBD regions over the genome of individuals (I) with 1st cousin pedigree I first cousins (1C) All over the genome IBD comes in segments Specific to each individual
- 9. Pedigree (fg) provides a value averaged over all possible lines of descent: E(f) = fg Large variability of the genome-based IBD around fg 1C: 95% interval = [0.02; 0.12] vs. fg=0.0625 fg=0.125 fg=0.0625 fg=0.0156 fg=0.0039 f Pedigree- vs Genome-based IBD
- 10. As pedigree relatedness becomes more remote, IBD over the genome might not exist anymore No IBD segments in 29% of individuals for 3C Given that there is IBD on the genome (f>0), IBD segments are long regardless of relatedness but they become rarer as relatedness becomes more remote (3C, 4C) fg Prob. of no IBD segments Average size of IBD segment # of IBD segments (f > 0) 1C 0.0625 <10-4 15.28 cM 14.75 2C 0.0156 0.01 11.70 cM 4.94 3C 0.0039 0.29 9.52 cM 2.06 4C 0.0010 0.67 7.41 cM 1.36 Pedigree- vs Genome-based IBD
- 11. IBD estimation from marker genotypes ⇒ need models to infer IBD from genotypes at markers over the genome If markers were fully informative, then identical alleles = identical by descent, IBD But markers have a limited number of alleles (SNPs only 2), so alleles may be identical (by state, IBS) without being IBD
- 12. Models for IBD inference To decide between IBD and identity by state (IBS): • Rely on allele frequency at each marker independently » rare alleles are more likely to be IBD than frequent alleles » Purcell et al (AJHG, 2007) (PLINK --het) → Inbreeding coefficient » Yang et al (AJHG, 2011) (GCTA) → Inbreeding & kinship coefficients
- 13. Models for IBD inference To decide between IBD and identity by state (IBS): • Rely on allele frequency at each marker independently » rare alleles are more likely to be IBD than frequent alleles • Rely on the segmental nature of IBD » stretches of markers with shared alleles are more likely to be IBD than isolated ones Need to define a minimum length threshold In humans usually around 1Mb, but still under debate (Pemberton et al, AJHG, 2012) » alleles shared within individuals (runs of homozygosity, ROH) McQuillan et al (AJHG, 2008) (PLINK --homozyg) » alleles shared between individuals Gusev et al (Genom Res, 2009) (Germline)
- 14. Models for IBD inference To decide between IBD and identity by state (IBS): • Rely on allele frequency at each marker independently » rare alleles are more likely to be IBD than frequent alleles • Rely on the segmental nature of IBD » stretches of markers with shared alleles are more likely to be IBD than isolated ones Need to define a minimum length threshold In humans usually around 1Mb, but still under debate (Pemberton et al, AJHG, 2012) » alleles shared within individuals (runs of homozygosity, ROH) McQuillan et al (AJHG, 2008) (PLINK --homozyg) » alleles shared between individuals Gusev et al (Genom Res, 2009) (Germline) ROHs > 1Mb
- 15. Models for IBD inference • Rely on both allele frequency and segmental nature of IBD, i.e. rely on haplotype frequency → hidden Markov model along the genome • FSuite/FEstim (Leutenegger et al, 2003; Gazal et al, 2014) »Probability of observing the marker genotypes written as a function of f inbreeding coefficient of the individual a length of IBD segments on the genome →Estimation of (f,a) by maximum likelihood & posterior IBD probabilities » Estimate of f depends on marker map density and allele frequencies Leutenegger et al, AJHG, 2003; Purcell et al, AJHG, 2007; Thompson, TPB, 2008; Browning & Browning, Genet, 2013; Han & Abney, EJHG, 2013; Gazal et al, Bioinformatics, 2014 Genotypes IBD
- 16. • Single-point: variance higher than other methods • ROHs: depend on the threshold • HMM: lowest bias and variance Comparison of methods Simulations with WTCCC haplotypes, Affymetrix 6.0 (517k SNPs) Gazal et al, Hum Hered, 2014 Single-point ROHs HMM on sparse map HMM modeling LD
- 17. FSuite Sans connaitre la généalogie: - Détecter et estimer f par individu - Inférer le type d’apparentement entre les parents d’un individu 1C; 2C; 2x1C; AV. - Estimer la proportion de ce type de mariage dans une population 1C: cousins germains; 2C: cousins de 2nd degré 2x1C: double cousins germains; AV: oncle/nièce;
- 18. Final Phase African (AFR,7) 661 African Caribbean in Barbados (ACB) 96 African Ancestry in Southwest United States (ASW) 61 Esan in Nigeria (ESN) 99 Gambian in Western Division, The Gambia (GWD) 113 Luhya in Webuye, Kenya (LWK) 99 Mende in Sierra Leone (MSL) 85 Yoruba in Ibadan, Nigeria (YRI) 108 European (EUR,5) 503 Utah residents with European ancestry (CEU) 99 Finnish in Finland (FIN) 99 British in England and Scotland (GBR) 91 Iberian populations in Spain (IBS) 107 Toscani in Italy (TSI) 107 East Asian (EAS,5) 504 Chinese Dai in Xishuangbanna, China (CDX) 93 Han Chinese in Bejing, China (CHB) 103 Southern Han Chinese, China (CHS) 105 Japanese in Tokyo, Japan (JPT) 104 Kinh in Ho Chi Minh City, Vietnam (KHV) 99 South Asian (SAS,5) 489 Bengali in Bangladesh (BEB) 86 Gujarati Indian in Houston,Texas (GIH) 103 Indian Telugu in the United Kingdom (ITU) 102 Punjabi in Lahore, Pakistan (PJL) 96 Sri Lankan Tamil in the United Kingdom (STU) 102 Admixed American (ADM,4) 347 Colombian in Medellin, Colombia (CLM) 94 Mexican Ancestry in Los Angeles, California (MXL) 64 Peruvian in Lima, Peru (PEL) 85 Puerto Rican in Puerto Rico (PUR) 104 TOTAL 2,504 • 2 504 individus de 26 populations (5 régions) • Présence de populations métissées • FSuite est utilisé sur chaque population • Calcul des fréquences alléliques par population (Freq. SAMPLE) • Filtrage sur les polymorphismes fréquents (MAF>5%) : 81M -> 3M Variants Application au panel 1000 Génomes
- 19. Consanguinité génomique dans le panel 1000 Génomes 595 individus sur les 2504 sont inférés consanguins (24% du panel), • essentiellement en Asie du Sud (SAS, 45%) ou en Amérique (AMR, 41%). Présence de consanguinité dans toutes les populations à des fréquences différentes : • plus de 25% d’individus consanguins dans 11 populations; • moins de 5% d’individus consanguins dans 6 populations; 1/16 = 1/16ème du génome homozygote par descendance (6,25% HBD) 2x1C: double cousins germains; AV: oncle/nièce; 1C: cousins germains; 2C: cousins 2nd degré
- 20. Deux-tiers de ces individus consanguins (64/94) proviennent de 3 populations d’Asie du Sud : ITU (10), PJL (22), STU (32). Consanguinité génomique dans le panel 1000 Génomes 2x1C: double cousins germains; AV: oncle/nièce; 1C: cousins germains; 2C: cousins 2nd degré Consanguinité éloignée : 501 individus inférés comme issus de couples de cousins au 2nd degré (2C) La population finlandaise (FIN), a un tiers des individus détectés consanguins (34%, tous 2C). Ceci est en accord avec l'histoire de la population finlandaise : un petit nombre de fondateurs et très peu d’immigration. Consanguinité proche: 94 individus inférés comme issus de couples de cousins germains (1C) ou plus proche
- 21. TGP2457 Panel sans apparentement, ni consanguinité très proche (1er et 2nd degrés) TGP2261 Panel sans apparentement, ni consanguinité proche (1er, 2nd et 3ème degrés) supplément Table S4 (liste des 2504 individus)
- 22. Outline 1. Consanguineous populations 1. Definition and characteristics 2. Measures of consanguinity 3. Consanguinity and Human Genetic Reference Panels 2. Human disease investigation in consanguineous populations 1. Rare monogenic diseases 2. Common multifactorial diseases
- 23. Fom Manolio et al, Nature 2009 Human disease investigation • Rare monogenic diseases »More than a third (3,000) of Mendelian disorders with an identified gene (Bamshad et al, 2011) • Common multifactorial diseases »Thousands of common variants with small/modest effects that cannot explain disease heritability → interest in rarer variants (<5%) via next generation sequencing (NGS) Linkage analysis Association analysis GWAS
- 24. Interesting characteristics in consanguineous populations for disease investigation • Identification of mutations involved in rare monogenic diseases » Within IBD recessive diseases (homozygosity mapping) » Between IBD in population isolates dominant diseases • Identification of variants involved in common multifactorial diseases » Within IBD variants with recessive effects » Population isolates Few founding individuals, within/between IBD reduced genetic complexity Shared environment reduced environmental heterogeneity
- 25. Outline 1. Consanguineous populations 1. Definition and characteristics 2. Measures of consanguinity 3. Consanguinity and Human Genetic Reference Panels 2. Human disease investigation in consanguineous populations 1. Rare monogenic diseases → recessive diseases 2. Common multifactorial diseases
- 26. Homozygosity mapping Lander and Botstein, Science, 1987 • Method to localize genes involved in rare recessive diseases • An inbred affected is likely to receive two disease alleles IBD »IBD region around the disease locus • Search the genome for a region where independent inbred affecteds are IBD • Measure the evidence for linkage in the region with a LOD score patient 1 patient 2 patient 3 patient 4 shared IBD region among patients
- 27. Powerful approach • Even a single patient is informative for disease gene location (linkage), not true in outbred population • A patient, offspring of first cousins, is as informative as an outbred nuclear family with 3 affected siblings • Very interesting as multiple affected siblings are unlikely for recessive diseases (1 out 4 expected) But it requires knowing the pedigree of the patients
- 28. Genomically controlled homozygosity mapping Leutenegger et al, AJHG, 2006 • H1: disease locus at position k H0: position k is a random point of the individual’s genome • Y set of marker genotypes → observed X set of marker IBD statuses (0 or 1) → unobserved For each affected individual: estimation of f, computation of the posterior IBD probabilities at each marker k [P(Xk|Y)] q: disease allele frequency; genetic model: fully penetrant recessive disease • For a sample of independent affected inbred individuals: FLOD = Σi FLODi ffqf YXqPYXP FLOD q kk i 1 log )1( )0()1( log 0→≈ −+ =+= = ).( ).( log 0 1 HdataobsP HdataobsP =Homozygosity mapping LOD score = HMLOD
- 29. Genomically controlled homozygosity mapping Simulation of 2x1C offspring – Type 1 error rate • HMLOD with incomplete pedigree: high increase in type I error rate for standard homozygosity mapping when inbreeding is underestimated • FLOD and HMLOD with complete pedigree: similar type I error rate → FLOD controls well for the presence of inbreeding x4 HMLOD complete pedigree (2x1C) FLOD without pedigree HMLOD incomplete pedigree TypeIerror 2x1C
- 30. Taybi-Linder syndrome / MOPD1 • Sample: 3 patients with pedigree information, 1 without » All patients shown to be inbred FEstim
- 31. • Sample: 3 patients with pedigree information, 1 without » All patients shown to be inbred • Candidate region on chr 2q14 » Using the 3 patients with pedigree information, no significant HMLOD » After including patient 4, candidate region with FLod=3.28 FEstim FLod Taybi-Linder syndrome / MOPD1
- 32. • Sample: 3 patients with pedigree information, 1 without » All patients shown to be inbred • Candidate region on chr 2q14 » Using the 3 patients with pedigree information, no significant HMLOD » After including patient 4, candidate region with FLod=3.28 • Complete sequencing (NGS) of the candidate region → Identification of a gene coding for a small nuclear RNA (U4atac) of the minor spliceosome FEstim FLod Edery et al, Science, 2011 Taybi-Linder syndrome / MOPD1
- 33. • Sample: 3 patients with pedigree information, 1 without » All patients shown to be inbred • Candidate region on chr 2q14 » Using the 3 patients with pedigree information, no significant HMLOD » After including patient 4, candidate region with FLod=3.28 • Complete sequencing (NGS) of the candidate region → Identification of a gene coding for a small nuclear RNA (U4atac) of the minor spliceosome FEstim FLod Edery et al, Science, 2011 This would not have been found by exome sequencing. The first step of homozygosity mapping was essential. Taybi-Linder syndrome / MOPD1
- 34. Outline 1. Consanguineous populations 1. Definition and characteristics 2. Measures of consanguinity 2. Human disease investigation in consanguineous populations 1. Rare monogenic diseases 2. Common multifactorial diseases 1. Focus on inbred cases from GWAS case-control data 2. Focus on isolated populations
- 35. • Genome-wide case-control association studies (GWAS) have identified many common variants over the past 10 yrs »Single marker tests, additive genetic models • GWAS were not designed to detect rare variants or variants with recessive effects • Can focusing on inbred cases identified in GWAS data help identify such variants? →HBD-GWAS strategy (Genin et al, Hum Hered, 2012) »Detect inbred cases in GWAS case-control data »Perform genomically-controlled homozygosity mapping Focusing on inbred cases in GWAS
- 36. WTCCC Type 2 Diabetes •1,924 cases / 2,938 controls »Affymetrix 500k SNPs •17 cases found to be inbred with f>0.01 •5 cases shared an IBD segment on chr 1, no controls »Only one gene (NEGR1), previously found associated with obesity FLOD≥1
- 37. Outline 1. Consanguineous populations 1. Definition and characteristics 2. Measures of consanguinity 2. Human disease investigation in consanguineous populations 1. Rare monogenic diseases 2. Common multifactorial diseases 1. Focus on inbred cases from GWAS case-control data 2. Focus on isolated populations
- 38. Focusing on isolated populations … To study complex traits : Reduced genetic, phenotypic and environmental heterogeneity … To study rare variants : Each population isolate is a random draw from the general population → some rare variants will become more frequent in the isolated population → more power to identify them
- 39. Next-generation association study in isolated populations Cost efficient strategy originally proposed in the Icelandic population »SNP genotyping on the whole population (scaffold) »Focus on a subset of the population for sequencing »Impute back the detected variants in the whole population »Perform association analysis Zeggini, Nat Genet, 2011 Holm et al, Nat Genet, 2011
- 40. Next-generation association study in isolated populations Zeggini, Nat Genet, 2011 Holm et al, Nat Genet, 2011 Cost efficient strategy originally proposed in the Icelandic population »SNP genotyping on the whole population (scaffold) »Focus on a subset of the population for sequencing »Impute back the detected variants in the whole population »Perform association analysis
- 41. How to select the subset to be sequenced ? Need to select “representative” individuals Each selected individuals should be • Representative of many other individuals • Not representative of another selected individuals Representative = sharing genome IBD
- 42. Measure of representativeness for selection • Kinship coefficient between individuals = proportion of genome shared IBD → a measure at the genome-level Abecasis et al, 2010 (ExomePicks); Urrichio et al, Genet Epi, 2012 • Use the shared IBD segments between individuals to measure the portion of the population genome that each selected individual provides (utility) Gusev et al, 2009 (Germline); Gusev et al, Genet, 2012 (INFOSTIP) Select individual A IBD segments
- 43. Cilento population isolate • Ancient origin: ~10th century • Bottleneck Plague in 1656 • Geographical isolation • Population size ~2,000 inhabitants • Genealogical data: From civil census and church archives going back to 1600 • Phenotype data for more than 400 traits • Genotype data on Illumina SNP chip for 1,444 individuals
- 44. Cilento population isolate • Sequencing 250 individuals will allow to capture > 80% of the population genome • After having selected 1,153 individuals, additional individuals only brings their own genome
- 45. Conclusion • Consanguineous populations have interesting characteristics for human disease investigation: IBD within (consanguinity) and between (kinship) individuals » rare monogenic diseases » common multifactorial diseases • Population isolates and rare variants involved in common multifactorial diseases • Need to keep developing methods for accurate IBD estimation from the genome
- 46. Acknowledgements Inserm, France Anne-Louise Leutenegger Steven Gazal Marie-Claude Babron Céline Bellenguez Hervé Perdry Emmanuelle Génin Françoise Clerget-Darpoux TALS study Patrick Edery, CHU Lyon IGB-CNR, Naples Teresa Nutile Marina Ciullo CEPH, Paris Jean François Deleuze