AI summary by TwinMind
Lecture 9: Machine-learning Approach
MIT OpenCourseWare
1h 28m 5s
Introduction to Genetics and Genomics
- Basic genetics evolution to genomic area by increasing size, scope, quality of genetic studies
- Origin: Human Genome Project first draft publication
- Future agenda item: SNP map promises to revolutionize disease mapping and human history tracking
- SNPs (Single Nucleotide Polymorphisms): subtle variations in human genome across individuals
- Humans 99.9% genetically similar, but subtle one-base variations make differences
- Billions spent on Human Genome Project for tracking genetic disease causes
Historical Context and Foundations
- 1980s: Human genome project development began
- David Botstein 1977: showed natural markers (polymorphisms) on genome make individuals different
- Central dogma: DNA → RNA → Proteins → Observable traits (disease susceptibility, physiology, metabolism, drug resistance)
- Polymorphisms as natural markers created intellectual foundation for Human Genome Project
Single Gene Diseases
- 400+ single gene diseases identified
- Classifications:
- Dominant vs recessive inheritance
- Autosomal vs X-linked
- Examples: Huntington disease, cystic fibrosis, hemophilia
- Simple mutations: base changes from normal coding to stop sequences
- Queen Victoria hemophilia case study: family tree analysis across generations
- Demonstrated Mendelian inheritance patterns
- Historical impact on European monarchy and Russian Revolution
Genetic Terminology and Concepts
- Alleles: Two copies per gene (one from mother, one from father)
- Locus: Physical location on chromosome
- Linkage: Proximity of alleles on chromosome
- Marker: Trackable allele at specific position
- Physical distance: Number of bases between points
- Genetic distance: Probability of recombination (measured in centimorgans)
- Phenotype: Observable characteristics
- Genotype: Internal DNA-coded information
- Penetrance: Probability of developing phenotype given specific allele
Genetic Distance and Recombination
- 1 centimorgan = 1% recombination probability ≈ 1 megabase
- Human autosomal physical map: 3 billion bases
- Linkage map differences: 2800 centimorgans (male) vs 4200 centimorgans (female)
- Female genetic code larger due to second X chromosome
- Co-segregation: Alleles transmitted together to next generation
Hardy-Weinberg Equilibrium
- Fundamental quantitative law: p² + 2pq + q² = 1
- Predicts distribution of major/minor alleles and heterozygotes
- Assumptions: random mating, infinite population
- Cystic fibrosis example: 1/2500 affected → 1/25 carriers in Caucasians
- Deviations indicate population stratification or selection pressure
Population Genetics Examples
- Blood group study (1975): Hardy-Weinberg predictions closely matched observations
- Sickle cell anemia in West Africa: excess heterozygotes due to malaria protection
- Demonstrates balancing selection maintaining allele frequencies
Complex Traits and Diseases
- Most common diseases don't follow single-gene models
- Complex traits involve:
- Multiple genes interaction
- Gene-environment interactions
- Incomplete penetrance
- Phenotypic variability
- Examples: BRCA1 breast cancer risk varies by age, retinitis pigmentosa (14 independent mutations)
Study Design Methods
Linkage Analysis
- Traditional pedigree analysis
- Likelihood ratio/LOD scores comparing models
- Large pedigrees required
- Multiple comparisons problem
Allele Sharing Methods
- Sibling-based studies
- Nonparametric approach
- Tests deviation from expected random distribution
- Requires large sample sizes
Association Studies
- Case-control design
- Population stratification concerns (Finland vs Sicily example)
- Parametric statistical methods
Transmission Disequilibrium Test (TDT)
- Uses parent-child trios
- Parents as internal controls
- Eliminates population stratification
- Child affected, parents unaffected
- Challenges: finding parents, phenotypic stratification
Quantitative Trait Loci (QTL)
- Continuous rather than binary traits
- Survival analysis approaches
- Breast development study example
- Animal models often required
- Censored data complications
Large-Scale Phenotypic Studies
- Framingham Heart Study: 50,000 people, 3 generations
- Nurses' Health Study: 150,000 women, 30+ years follow-up
- Comprehensive phenotypic data collection
- Genotyping costs: ~$0.45/SNP, ~$90,000/individual full map
- Cost barrier: $90 million for 1000 individuals
Haplotype Mapping and Linkage Disequilibrium
- Haplotypes: chromosome segments inherited together
- Linkage disequilibrium blocks reduce genotyping requirements
- Example: 59 SNPs reduced to 5 tag SNPs without information loss
- HapMap project: systematic genotyping at 30kb resolution
- 30 trios for phase determination
Chromosome Structure and Evolution
- Genome organized in stable blocks
- Crohn's disease region: 11 sub-regions from 4 ancestral haplotypes
- Exponential savings with block length
- Average 10% genotyping requirement with optimal tagging
Key Insights for Medical Applications
- Good phenotype characterization critical for success
- Hypothesis-free genome-wide approaches now possible
- Technology costs decreasing, phenotyping remains expensive
- Systematic genome analysis provides stronger evidence than single-gene studies
- Meta-analyses will consolidate findings across studies
