2019-10-04

MBP Tech Talks

Fall 2019

Fundamentals of Genome Sequencing and Applications

Motivations

  • Lots of research involves genome sequencing or and analysis
  • Few resources covering the fundamentals
  • Information is often scattered through blog posts, presentations
  • Multidisciplinary - difficult for a single person to discuss thoroughly

Goals for the fall semester

  • Understand foundations of genome sequencing
    • How DNA is captured and read
    • What makes good sequencing data
  • Sequence alignment
  • Wet and dry sides of sequencing experiments
  • How to think about good experiment design when sequencing

Outline

Part 1: Fundamentals of genome sequencing

  • History of reading DNA
  • Massively-parallel sequencing
  • Standard file formats
  • Sequence alignment
  • Alignment metrics

Outline

Part 2: Applications of genome sequencing

  • Mutation detection
  • Chromatin accessibililty
  • Histone modifications and protein binding
  • Transcriptome sequencing
  • Chromatin organization
  • DNA methylation

Session structure

  • HSB 100, Friday 12:00 - 14:00
  • No food or drinks
  • Each session has 2 parts, break in the middle

Fundamentals of DNA sequencing

DNA molecules

Alberts, Molecular biology of the cell, 6ed. pg. 176

  • DNA is double-stranded polymer
  • DNA contains 4 nucleotides: adenine (A), cytosine (C), guanine (G), thymine (T)
  • DNA has a sugar-phosphate backbone
  • Nucleotides come in complementary pairs: A-T, C-G

DNA molecules

Alberts, Molecular biology of the cell, 6ed. pg. 177

  • DNA has a right-handed orientation
  • Strands are oriented in opposite directions
  • We measure direction by counting 5’ to 3’

DNA encodes information for organisms

Alberts, Molecular biology of the cell, 6ed. pg. 178

  • Complementary strands allow for replication
  • Sequences themselves code for proteins inside cells

DNA sequencing

  • Genome: set of all DNA inside an organism
  • Sequencing: the process of measuring the order of these nucleotides in a set of cells

Sanger sequencing

  • Just from a high level, since details aren’t totally relevant, just the idea
  • Gel electrophoresis
    • Small fragments of DNA are loaded into lanes of gel
    • Small electric potential pulls DNA through gel
    • Larger resistance on longer fragments
  • Load polymerase, primers, and ddNTPs to template strand
    • Primers to act as substrate for polymerase
    • Polymerase to bind ddNTPs to template strand
    • Add ddNTPs one at a time to control which nucleotide gets added next

Sanger sequencing

Mardis, 2013

  • Separate small portion after adding a particular ddNTP
    • Keep all portions after adding a given ddNTP together
    • Load these in a lane
  • Bright band in the lane gives nucleotide
  • Vertical position gives order in sequence

Sanger sequencing

Pros

  • Sequence DNA fragments
  • Can visually see the order of nucleotides

Cons

  • Limited by sharpness of bands
  • Limited by length of gel
  • X-ray gel exposure, image development, one nucleotide at a time (laborious)

Sequencing by synthesis

  • Use ideas of Sanger sequencing:
    • Primers, polymerase, nucleotides
    • Synthesize complementary strand base-by-base
  • Instead of measuring all at once via gel, measure fluorescence

Sequencing by synthesis

Sequencing by synthesis

Sequencing by synthesis

Sequencing by synthesis

Sequencing by synthesis

Sequencing by synthesis

Massively parallel sequencing

Massively parallel sequencing

Massively parallel sequencing

Massively parallel sequencing

Massively parallel sequencing

Massively parallel sequencing

Massively parallel sequencing

Massively parallel sequencing

Massively parallel sequencing

Massively parallel sequencing

Massively parallel sequencing

Massively parallel sequencing

  • Complement of captured bases to reveal the original DNA sequence
  • Image of the entire slide allows simultaneous capturing of millions of fragments of DNA

Storing DNA sequences

FASTA format

  • Sequence name
  • Sequence string
>Name
SEQUENCE
>reference-seq-name1
ATCTATACTTTATCTTTATCTTTA
>reference-seq-name2
ATTTTATCGCGTAGCTAGCTGGCT
  • Pairs of lines are the fundamental units
  • Everything encoded in plain text

FASTA format Pros

  • Simple to understand
  • Encodes DNA in plain letters
  • Easy to parse, edit, etc

FASTA format Cons

  • Assumes sequences are exact
    • Often not true in practice
  • Often extremely large files
    • Human genome: 3 Gbp
    • Tend to be gzipped to compress size
  • No computational optimization
    • Random access
    • ASCII encoding is overkill when you only need 2 bits

FASTQ format

  • Very similar to FASTA
  • Sequence name
  • Sequence string
  • Description
  • Quality scores (uncertainty in measurement)
  • Analogous to working with sig figs, % err in measurements

FASTQ format

Quality scores

  • Consider same set of reads as before
  • Fragments spread out on a substrate
  • Ideally we’d be able to measure a single strand of DNA
  • More robust measurements come from having clones of the same fragment

FASTQ format

Quality scores

FASTQ format

Quality scores

  • Consensus call is blue
  • Not unanimous

\(p = \mathbb{P}[\text{incorrect call}] = \frac{1}{9}\)

FASTQ format

Quality scores

\(p = \mathbb{P}[\text{incorrect call}] = \frac{3}{9} = \frac{1}{3}\)

  • This is a simple model for calculating errors in calls
  • More complicated methods can be used to calculate \(p\), based on the chemistry, the asymmetric errors, input sample base distribution, etc

FASTQ format

Quality scores

  • Modern sequencers are very accurate
  • Often \(p \approx 0\)
  • Phred quality score \(q = -10\log_{10}(p)\)
  • \(q \in [0, \infty)\),
  • \(q = 10 \implies p = \frac{1}{10}\)
  • \(q = 20 \implies p = \frac{1}{100}\)
  • \(q = 30 \implies p = \frac{1}{1000}\)
  • \(q = 40 \implies p = \frac{1}{10000}\)

FASTQ format

Quality scores

  • Encode \(q\) as a single ASCII character
  • ASCII(round(\(q\)) + 33)
  • \(q = 0 \implies\) !
  • \(q = 40 \implies\) I
@reference-seq-name1
ATCTATACTTTATCTTTATCTTTA
+
GFFFFBBBCBCCBBAAA:::;;;;
@reference-seq-name2
ATTTTATCGCGTAGCTAGCTGGCT
+
FFFEEBBBCBCCBBAAA:::;;;;

DNA sequencing data is full of errors

  • Say you have 1000 fragments, each 100 bp long, each base has \(q = 40\)
  • What is the probability you have no errors anywhere in your data?

DNA sequencing data is full of errors

  • \(\mathbb{P}[\text{no errors}]\)
  • \((\mathbb{P}[\text{no errors in read}])^{1000}\)
  • \(((1-\mathbb{P}[\text{incorrect base call}])^{100})^{1000}\)

  • \((1-p)^{100000} \approx 4.54 \cdot 10^{-5}\)

  • \(\mathbb{E}[\text{incorrect base calls}] \approx np = 100000 \cdot \frac{1}{10000} = 10\)

DNA sequencing data is full of errors

Sequencing is a random sampling problem

  • Many steps involve complicated chemistry
  • DNA doesn’t come in these pre-defined chunks that we can measure
  • All sequencing is random sample of input space
  • Similar to taking polls of demographics to extrapolate to the population
  • Prime example of Hidden Markov Model
  • Many steps in sequencing are stochastic
    • Composition of input DNA
    • Fragmentation
    • Adapter ligation
    • Annealing to substrate
    • Amplification
    • Errors in base calls
    • Trillions of possible fragments, only millions sequenced
  • Single measurements are not deterministic

Break

Assessing DNA sequencing data quality with FastQC

FastQC

  • Tool to assess quality of sequencing data
  • Produces HTML report
fastqc {FASTQ}

FastQC Examples

Summary

  • A brief history of DNA sequencing and various methods
  • Phred quality scores
  • Storing DNA sequences (FASTA, FASTQ)
  • FASTQ quality control metrics
  • Sequencing as a random sampling measurement

What I didn’t cover

  • Alternative sequencing technologies (PacBio, Oxford Nanopore)
  • Chemistry of these technologies
  • What to do with these data

Next time

  • Exact sequencing alignment
  • String alignment to a reference
  • Naive alignment
  • Boyer-Moore algorithm