May 4th, 2018

DNA sequencing & assembly

How to sequence your own DNA

Sanger method

  • Several DNA copies are cut using 4 restriction enzymes
  • They cut specifically on each type of nucleotide
  • They attach a fluorescent marker to the end

The detector sees different colors

Fragments are separated by electrophoresis

Fragment length tells us the position of the nucleotide

The result is called chromatogram

From it we get the letters of the sequence

Shotgun DNA Sequencing

Shotgun DNA Sequencing (2)

Technology changes fast

In 2001, the cost of sequencing the first human genome was USD 108

Today you can have your own genome for 1000 USD

The problem is no longer how to do the experiment

Instead is how do we make sense of the results

DNA sequencing is cheap

A DNA sequencer in every desktop

First computers where big and expensive

Only in a few universities, used by experts

Then there was one on every office… and home

Today everybody has one… in the pocket

A PlayStation has more power than the biggest computer of 1998

Can the same happen with DNA sequencing?

The next iPhone

Today you can buy a DNA sequencer of the size of an iPhone

… at the price of an iPhone

Next step: people will make apps for DNA sequencer

Computing power grows exponentially

Moore’s Law

Big Picture

  • Sequencing cost is going down
  • Moore’s law: computing is cheap

  • There is a phase transition: we changed from “solid” to “liquid”

  • Rules are changing very fast

  • For example, patents are obsolete

Producing data is cheap

There is already a lot of public data

  • Tara Oceans published 7.8 Terabytes of metagenomic data
    • equivalent to 1.660 DVDs
  • Anybody can discover new knowledge there
    • You can do it!
    • And also many other people

This is a race

It does not depend on hands and wallets

It depends on brains and guts

All Science is Data Science

But Data Science is not about Data

Science is about obtaining

  • Information
  • Knowledge
  • Wisdom

Genome Assembly

The sequencer produces small pieces

Current technology allows us to read DNA in runs of ~100-600 letters. Imagine a book of 1000 pages:

  • several copies of the book are cut randomly in one million pieces
  • different pieces may overlap
  • half of the pieces are lost
  • the remaining half is splashed with ink in the middle

The problem is to reconstruct the original book

In pictures

The protagonists of the story

  • Genome: Only one, length G, let’s say \(10^7\)
G <- 1E7
  • reads: N of them, let’s say \(10^4\), each of length L
N <- 1E4
L <- 300

In Pictures

Reads are randomly taken from the genome

  • Each read has a start and end position on the genome
  • L must be equal to end-start
  • We take start randomly, then we calculate end
start <- sample.int(G, size=N, replace=TRUE)
end <- start + L

Each read has start and end

Sometimes reads do overlap

Sometimes they do not overlap

How do we know when two reads overlap?

How do we know when two reads overlap?

This is an event. That is, a function that returns TRUE or FALSE

  • What are the inputs
  • When must it return TRUE?

What is the overlap size?

Negative gap size = overlap

If two reads do overlap, the overlap will be negative

Thus overlap_size is -gap_size

If we do not know their position, how can we detect if two reads do overlap?

Distribution of gap length

We only see overlaps over a threshold T

The assembler needs to compare all reads to all reads and see if they overlap.

  • Can we detect overlap of 1 nucleotide?
  • What is the best threshold T?

If T is too small we get wrong results

If we decide that two reads overlap only when they share 1 letter, we will put together 25% of them every time

The best T has to be big enough to guarantee that the reads do not overlap by chance

If the reads do not match by chance, then there is a biological reason

Bigger values of T reduce the probability of “overlap by chance”

If two reads overlap, they are in the same Contig

A group of contiguous sequences is called Contig

The goal of the assembly process is to find one contig.

The sequence of this contig will be the genome

Most of times we do get several contigs