AI summary by TwinMind
Lecture 7: Informational Resources
MIT OpenCourseWare
1h 15m 6s
Introduction and Central Dogma Overview
- Instructor at CHIP presenting biomedical information resources
- Focus on human genome study-related information
- Cover actual locations (mostly websites) for finding biomedical data
- Central dogma: DNA → RNA → proteins → phenotypes
- DNA in nucleus encodes information
- Transcribed to mRNA, translated to proteins
- Proteins responsible for external manifestations (metabolism, physiology, diversity, disease, drug response)
- Genotype: information encoded in DNA
- Phenotype: observable/measurable characteristics
Data Type Characteristics and Differences
- Genotype characteristics:
- Digital - exact representation using 4 symbols (A,T,G,C)
- Can represent 3 billion base pairs exactly
- Insertions, deletions possible
- Phenotype characteristics:
- Analog - collective in nature
- Cannot be measured/defined precisely
- Environmental factors affect measurements
- Proteins not uniquely determined by sequence
- Historical progression:
- Mendel's inherited traits study: 1866
- DNA discovery: same period, no known connection
- Proof genes made of DNA: 1952
- DNA structure/replication: 1953
- Genetic code deciphered: 1961-1966
- Introns discovered: 1977
- Human Genome Project completed: last year
Human Genetic Variation and Polymorphisms
- No two humans genetically identical
- Differences due to polymorphisms:
- Single nucleotide polymorphisms (SNPs)
- Microsatellite repeats
- Insertions, deletions, translocations
- Average: 1 polymorphism per 1000 bases
- 3 billion base pairs = very large number of differences
- Individual genomes don't match reference exactly
- Understanding consequences major current challenge
Computational Operations by Data Type
- DNA level operations:
- Sequence matching for gene discovery
- Gene finding in sequences (computationally complex)
- Homology searches across organisms
- SNP detection by comparing sequences
- RNA level operations:
- Alternative splicing analysis
- Transcription rearrangements
- Expression analysis, microarray clustering
- Protein level operations:
- Active domain prediction
- 3D structure prediction (very complex)
- Homology/conservation studies across organisms
- Metabolic/regulatory pathway construction
- Phenotype level operations:
- Population genetics
- Association studies (SNP-disease correlations)
- Clinical trials for validation
Gene Definition Context-Dependent
- Multiple definitions depending on perspective:
- Classical geneticist: smallest unit of inheritance
- Medical researcher: disease-causing trait
- Molecular biologist: recipe for protein(s)
- Biochemist: element in metabolic network
- Modern geneticist: functional locus on chromosome
- Bioinformatician: annotated DNA stretch in database
GenBank and Sequence Data
- GenBank characteristics:
- Largest repository of sequence data
- Direct submissions from researchers worldwide
- 22+ million sequences from 100,000+ organisms
- ~30 billion nucleotides total
- Part of NCBI (National Center for Biotechnology Information)
- Growth pattern:
- Very steep increase in base pairs and sequences
- Sequences getting longer over time
- Genome assembly process:
- Find overlapping sequences
- Reconstruct entire genomes
- Human genome: 5-6x coverage validation
- Multiple levels of coverage for validation
Completed Genomes
- 1000+ virus genomes
- Many bacterial genomes
- Eukaryotes hardest to sequence
- Human genome considered finished
- Chimp genome recently released (1% difference from human)
- New genomes weekly for viruses/bacteria
Genome Browsers and Visualization
- Golden Path:
- Genome browser for multiple organisms (human, mouse, rat, chimp, etc.)
- Graphical view of chromosomal regions
- Downloadable data in multiple formats
- Arbitrary DNA sequence retrieval
- Absolute positioning of genetic elements
- Multiple annotation tracks (genes, SNPs, conservation)
- Customizable display options
- NCBI Map Viewer:
- Alternative graphical genome browser
- Multiple map types: sequence, cytogenetic, linkage, homology
- More detailed but harder to use
- Extensive abbreviations and complex interface
SNP Databases and Resources
- SNPs importance:
- Most common form of genetic variation
- Genomic markers at fixed locations
- Causal candidates for diseases
- Evolutionary markers
- Population genetics applications
- dbSNP (NCBI):
- 4+ million human SNPs (likely 5+ million now)
- ~50% validated (observed multiple times independently)
- SNP Consortium Database (TSC):
- All SNPs validated
- Frequency data for different populations
- Baseline frequencies crucial for association studies
- HapMap Project:
- Building haplotype map of human genome
- Next step after SNP frequency determination
- Other databases:
- HGBase: manually curated, high quality, disease associations
- ALFRED: frequency data from isolated populations
- SNiPer: integrates information from multiple sources
Population Differences in SNP Frequencies
- Different populations have different SNP frequencies
- Example: African-American vs Caucasian populations show major/minor allele reversals
- Critical for association studies design
- Baseline population specification essential
Gene Information Resources
- LocusLink:
- Curated directory of genes from 13 organisms
- Establishes connections between sequence and descriptors
- Provides standardized gene nomenclature
- Links to other databases
- Gene naming complexity due to historical reasons
- UniGene:
- Automated clustering of GenBank sequences
- Groups sequences from same genomic region
- 38 organisms (rapidly growing)
- Represents multiple forms of same gene
- HomoloGene:
- Database of orthologs across 25 organisms
- 470,000 ortholog pairs
- Similarity-based matching
- Triplet confirmation for higher quality
- Partly automated, partly curated
Ensemble Genome Browser
- European Bioinformatics Institute resource
- Automated genome annotation system
- 10 organisms with extensive information
- Powerful query interface
- Own gene naming scheme (integration challenge)
- Combines multiple data types (genes, proteins, diseases, SNPs, microarrays)
Gene Regulation Complexity
- Extremely complex mechanism
- Multiple factors influence gene expression:
- Tissue type
- Developmental stage
- Time (circadian rhythm, long-term processes)
- External signals/stimuli
- Other gene expression states
- Transcription factors:
- Proteins binding to upstream gene regions
- Control gene expression
- Bind to promoter regions
- Form complexes for transcription activation
- Act combinatorially (not individually)
- Pattern recognition crucial for understanding
Transcription Factor Binding Sites
- Small DNA stretches (5-25 base pairs)
- Recognized by specific transcription factors
- Binding site proximity indicates factor interaction
- Detection challenges:
- Computationally difficult
- Consensus sequences not very specific
- Pattern matching methods required
- TRANSFAC database:
- Largest transcription factor database
- Experimentally validated binding sites
- Factor information and gene interactions
- Training data for computational methods
Microarray Data Resources
- GEO (Gene Expression Omnibus):
- 12,000+ hybridization experiments
- 500+ different platforms
- Powerful search interface for large datasets
- Stanford Microarray Database:
- Repository of Stanford experiments
- Portion publicly available
- NCI-60:
- 60 human cancer cell lines
- Gene expression profiles
- Drug activity correlation data
- PGA Projects:
- TREX: 565 arrays from mouse/rat disease models
- HOPG: 500+ arrays from human diseases
- Cardiogenomics: cardiac development data
- Human Gene Expression Index
Protein Databases Complexity
- More complex than DNA/RNA databases
- Proteins interact in 3D, catalyze reactions
- Harder to describe quantitatively
- Characteristics:
- Older databases
- Less integrated
- Less complete
- Nomenclature problems
- Less standardized
Major Protein Resources
- Swiss-Prot:
- 120,000 sequence entries
- 9,000 human proteins (small fraction of total)
- Core data: sequence, references, taxonomy
- Annotations: functions, domains, structures, diseases
- Own protein identifiers (linking challenges)
- PDB (Protein Data Bank):
- 3D protein structures
- Visualization options
- MMDB:
- NCBI implementation of structure database
- Pfam:
- Protein domains and families database
- Domain similarity measures
- Hidden Markov models
- Curated high-quality portion
- Automated lower-quality portion
Phenotype and Disease Resources
- OMIM (Online Mendelian Inheritance in Man):
- Catalog of human genes and genetic disorders
- 14,000+ entries (likely more now)
- Text articles about genes and disorders
- Cross-linked gene-disease information
- Biochemical features, mapping, clinical correlations
- Allelic variants with clinical outcomes
- PubMed:
- 12+ million biomedical literature citations
- Mid-1960s to present
- Links to full-text articles and supplementary materials
- 30 million searches per month
Gene Ontology
- Dynamic controlled vocabulary for biological concepts
- Standardized nomenclature for three domains:
- Molecular function
- Biological process
- Cellular component
- Hierarchical taxonomies with concepts/sub-concepts
- Work in progress, incomplete
- Challenges in formalizing natural language concepts
- Enables precise term definition and reference
Challenges and Future Directions
- Drowning in data, converting to knowledge difficult
- Need automated tools for:
- Data access
- Data interpretation
- Format conversion
- Biomedical data spans entire spectrum of knowledge representation
- Management techniques require diverse approaches
- Integration across different data types and naming schemes ongoing challenge
