Human Genome Project

1
Human Genome Project

• In early 2001, it was announced that the entire
  human genome had been sequenced
• Lets pretend this was true
• It wasnt really true, it was more of a publicity
  stunt
• It was a very rough draft
• Even today, 5 years later, it is not quite done
  although it is much closer

2
Human Genome Project

• What does the sequencing of the human genome
  mean?
• We now have 24 very long strings of letters (only
  A, T, C, G) representing the human genome
• These strings range from 55Mb to 220Mb, for a
  total of 3.2Gb
• How big is this?
• If the entire human genome were printed in the
  same font used by most encyclopedias, it would be
  10 times longer than the 2002 edition of
  Encyclopedia Britanica
• No information on variability or frequency
• Represents a single snapshot of the genome
• What good is it?

3
Genomics

• The general process of finding things within a
  sequence is known as Genome Annotation
• How can we identify coding genes in a sequence?

4
Finding Coding Genes

• Experimentally
• How does one find a gene?
• DNA ? mRNA ? Amino Acid/Protein
• Grab the mRNAs
• Use reverse transcription to create cDNA
• Look for the cDNA in the genomethis tells you
  where the gene is
• Use BLAST or a similar tool

5
Finding Coding Genes

• Complication Introns
• In the mRNA the introns have been removed, thus
  our cDNA also does not contain introns
• We need to find pieces of contiguous cDNA but
  should not expect to find the entire cDNA
  sequence as one large chunk
• What if we do find the entire cDNA?
• mRNA will occasionally be reverse transcribed and
  placed back in the genome, often at random points
• This is known as a Processed Pseudogene
• Its a pseudogene because while the complete
  coding sequence is present, there are usually no
  control regions
• In Eukaryotes, almost all intronless genes turn
  out to be pseudogenes

6
Finding Coding Genes

• We can find genes experimentallydoes this help
  us finding them computationally?
• What we really want is to find genes in a
  sequence de novo, not experimentally
• How might one do this?
• What would you look for?
• Known genes from other organisms?
• This would still require that the gene be
  identified experimentally in the other organism
• Does not help identify unique genes
• Start with a collection of genes already
  identified through experiments
• Identify commonalities

7
Finding Coding Genes

• Given a set of genes, find out what is common
  among them
• Search the full genome for these common features
• Include up- and downstream information, not just
  the coding sequence
• What are the common features of genes? What
  should we be looking for?

8
Prokaryotic Genes

• Prokaryotes are bacteria and similar unicellular
  organisms
• Prokaryotic genomes are simpler than eukaryotic
  genomes
• Smaller genomes
• High gene density, approximately 1 gene/kb on
  average
• No introns
• Little repetitive DNA

9
Prokaryotic Genes

• Open Reading Frame (ORF)
• A stretch of DNA bracketed by a start codon (ATG)
  and a stop codon (TGA, TAA, or TAG)
• The start codon can be tricky since ATG could be
  the beginning or the middle of a sequence (with
  another ATG upstream). But, the fact that
  another ATG is upstream does not automatically
  mean the second one is not the proper start codon
• Any stretch of DNA has six potential reading
  frames, so we have to examine all six for ORFs
• Are all ORFs coding genes?
• No, but all coding genes must contain an ORF?

10
Open Reading Frames

• Once weve found an ORF, how can we tell if it
  might be a a gene?
• Length
• How many codons are there?
• 64
• How many stop codons are there?
• 3
• What is the probability of any given codon being
  a stop codon?
• Approximately 1/21
• Thus for any random stretch of codons, on average
  one would expect a stop codon every 21 codons

11
Open Reading Frames

• Most prokaryotic genes are longer than 60 codons
• In E. coli the average gene is 316 codons long
• Statistically, if all codons appear with equal
  frequency in a random sequence, the probability
  that a stop codon will not occur over N codons is
  (61/64)N
• The ORF length at which we expect random sequence
  to have less than a 5 chance of not containing a
  stop codon happens to be N 60

12
Open Reading Frames

• Other features of ORFs Codon usage
• When we examine a lot of known genes, we find
  that some codons are quite common while others
  are rare
• Some amino acids are more common than others
• Even for the same amino acid, not all synonymous
  codons will occur with equal frequency
• Known as Codon Usage Bias
• This is true even when nucleotide biases are
  taken into account
• It probably has something to do with the
  efficiency of gene expression

13
Codon Usage Bias

• Codon usage biases tend to be species specific

Percentage of use of Serine codons in different
species

• One can calculate the proportion of codon usage
  for all known genes
• If a proposed ORF has similar proportions of
  codons as known genes, it is probably a real gene
• If it tends to have high proportions of rarely
  used codons, it may not be a real gene

14
Promoter Elements

• Promoters tell the transcription machinery (e.g.,
  RNA polymerase) where to start
• Are there common promoters?
• In E. coli the following consensus sequences are
  often found upstream of the start codon

• The better the -35 and -10 upstream region match
  a pair of these sequences, the higher the odds
  that it is a real gene

15
Promoter Elements

• Not every gene in prokaryotes has a promoter!
• Genes in prokaryotes often work in sets called
  Operons
• An operon consists of a single promoter followed
  by multiple coding sequences
• All of the coding sequences are transcribed
  simultaneously when the single promoter is
  activated
• Genes in operons usually work together as parts
  of a single system

16
Operons

• Operon Example Lactose Operon
• Three Genes
• Beta-galactosidase
• Lactose permease
• Lactose transacetylase
• Involved in the metabolism of lactose sugar
• Transcription creates one long mRNA, pieces of
  which are translated into three separate proteins

17
Operons

• How can you identify separate genes within one
  operon?
• There needs to be a sequence in front of the
  start codon for the ribosome to attach to
  (something other than a promoter)
• Known as Shine-Delgano sequences
• AGGAGGU is found in front most true start codons
  in mRNA

18
Operons

• Most operons have termination signals after the
  stop codons
• Inverted repeats CGGATGCATCCG
• This is identical on the reverse strand
• A stretch of 6 U follows the repeat
• These are found in gt 90 of prokaryotic operons
• These inverted repeats are 7 to 20 nucleotides
  long and usually very heavy in G and C

19
Prokaryotic Genes

• Complications
• Lateral gene transfer ( Horizontal gene
  transfer)
• When a species picks up a gene from another
  species rather than a direct ancestor
• Because different species have different GC
  contents and different codon usage patterns,
  lateral gene transfers will often appear very
  different from the other genes within the species
• This makes them easier to identify as foreign
  genesif you can identify them as genes in the
  first place!

20
Eukaryotic Genes

• Eukaryotes are much more complicated than
  prokaryotes
• Introns
• Break up the reading frame and make sequences
  much longer
• Tens of thousands of genes
• Lower gene density
• In worms and insects, 25 of the genome codes
  for proteins
• In humans it is closer to 3
• Lots of repetitive DNA
• Like searching for a needle in a haystack?
• A 2 gram needle in a 6kg haystack would be 1000
  times easier to find if genes were as different
  from the rest of the genome as a needle is to
  straw

21
Eukaryotic Genes

• Additional reading (available on website)
• Zhang (2002) Computational prediction of
  Eukaryotic protein-coding genes Nature Review
  Genetics 3698-709.
• This is a good overview of the techniques and
  problems involved with finding eukaryotic coding
  genes

22
Exons

• There are 5 types of exons
• Nonconding exons representing 5 and 3
  untranslated regions
• Initial exons containing the start codon
• Termination exons containing the stop codon
• Internal exons containing neither start nor stop
  codons
• Single exon genes, containing no introns and both
  start and stop codons
• In general, codons are more difficult to identify
  in eukaryotes because the reading frame (and even
  individual codons) is broken up by the introns

23
Exons

• Terminal (3) exons are easier to identify than
  initial (5) exons
• Many eukaryotic genes have AAUAAA or AAUUAAA
  tails some distance after the stop codon
• Promoters are harder to identify in eukaryotes
  than prokaryotes
• Many eurkaryotic genes have a TATA box
• Only 70 of human genes have this
• TATAWAW at position 25 (W indicates an A or a T)
• Initiator sequence (where transcription actually
  beginsnot the start codon)
• YYCARR with the A being the actual start of
  transcription
• Transcription factors (around 80)
• Most eukaryotes have CAAT (consensus is GCCAATCT)

24
Internal Exons and Introns

• Exons often end with GT and begin with AG
  (outside the actual coding sequence)
• At least 8 different kind of introns in
  Eukaryotes (i.e., at least 8 different ways that
  introns get spliced out)
• The most common intron is known as the GU-AG
• The first two nucleotides of the intron are GU
• The last two nucleotides of the intron are AG
• In yeast genes
• AGG100T100A100A100G100T10012PYNC100A100G100GT
• Introns must be at least 60 bases long (shorter
  introns cannot be recognized)
• Exons in vertebrates average 450bp, although some
  are lt 100bp and others are gt 2000bp

25
Introns

• The number of introns in a gene varies
  tremendously
• The 6,000 genes of yeast have a total of 239
  introns
• There are individual human genes with more than
  100 introns
• 95 of all human genes have at least 1 intron

26
GC content

• GC content does not vary as much from species to
  species in Eukaryotes as in Prokaryotes
• Within species, however, GC content can be
  structured into isochores
• Genes are associated with certain isochores
  isochores with high GC content have a much higher
  proportion of genes than isochores with low GC
  content
• Highest GC isochore in humans 3 of genome, but
  contains 80 of housekeeping genes
• Lowest GC isochore in humans 66 of genome, but
  contains 85 of tissue-specific genes
• L1 39 L2 42 H1 46 H2 49 H3
  54
• CpG islands
• Every known housekeeping gene is found in a CpG
  island
• A housekeeping gene is a gene expressed in all
  tissues at all times

27
Algorithms/Software

• Common ones are Grail EXP and GenScan
• No specifics will be discussed, but they
  generally come in three types
• Strict Rule-based
• Neural networks
• Hidden Markov Models
• Rules/training is species dependent e.g., a rule
  set derived for Drosophila may fail in humans
• However, comparative searches as already
  discussed (BLAST for Drosophila genes in humans)
  can help speed up the search for human genes to
  use in training algorithms

28
Algorithms/Software

• How good are the current approaches?
• So-so
• High false positive rate
• Identify genes that are not actually genes
• Fail to accurately predict complete genes
• Find some of the exons, but mess up the introns
• In 2000, the best predictor found true positives
  at about 40 accuracy and rejected true negative
  at about 30 accuracy
• Today, these numbers might be closer to 50
• One approach that helps is to use multiple
  programs/algorithms and look for consensus among
  their results
• If all of the programs identify something as a
  gene, there is a better chance that its correct
  than if only one found it

29
Algorithms/Software

• In the long run, even if a gene is identified
  with high confidence
• It needs to be confirmed experimentally
• Its function needs to be determined
• Function is most likely to be determined through
  experiments, but there are computational ways to
  make guesses
• For example, runs of 20-25 hydrophobic amino
  acids often indicate transmembrane domains

30
Non-coding elements

• What about things other than coding genes?
• 5 of the mammalian genome is functional
• Only 1-2 is actually coding sequence
• What about the other 3-4
• Noncoding RNAs
• Transcription factors and other regulatory sites
• How do we find these?
• Similar methods, for the most part

31
Non-coding RNAs

• Without the codon structure, they are trickier to
  find than coding genes
• They have promoters and transcription factors,
  but no start/stop codons or introns.
• They can also be very short

32
MicroRNAs (miRNAs)

• A large family of noncoding RNAs which help
  regulate gene activity
• They are only 21-22 nucleotides long
• How to identify them?
• The precursor is 70-100 nucleotides long with a
  lot of stem-loop structure
• Highly conserved among closely related species
• Diverge in a characteristic way changes tend to
  take place in a specific part of the stem
• Search for 100 nucleotides segments in a pair of
  species that differ by fewer than 15 of sites
  with fewer than 13 gaps
• Analyze each of these conserved regions for
  secondary RNA structure
• If the predicted structure looks correct, it is
  probably a miRNA

33
Other functional elements

• Primary approach is to look at multiple, closely
  related species and look for conserved upstream
  sequence
• searching for cis- acting elements
• trans- acting elements are essentially impossible
  to identify in this context
• Sequences which show few changes over time are
  presumably under selection and therefore must
  serve some function
• Perform multiple local alignments
• This approach is known as Functional or
  Phylogenetic Footprinting

34
Regulatory sequences

• Usually vaguely close to a gene, within a few kb
• However, they can be extremely short (e.g., just
  6 bases)
• They can be in any orientation
• Upstream/downstream
• Either direction on either strand
• Since finding any one potential regulatory
  sequence is easy to do by chance, it is better to
  look for multiple small sequences that are found
  near each other
• Look for word 1a spacer of some lengthword 2
• Use statistical analysis to determine if some
  combinations of words are found more often than
  expected by chance
• These are likely to have functional significance