• Everybody delivered
• Everybody said something interesting/original
• Everybody made different mistakes
• Together, you can have the “big picture”

I made a mix of all your answers so everybody can see the good and bad

Enables researchers to efficiently discover, validate, and screen genetic variants using a highly targeted approach

Supports multiplexing of hundreds to thousands of amplicons per reaction to achieve high coverage

Delivers highly targeted re-sequencing even in difficult-to-sequence areas, such as GC-rich regions

Allows flexibility for a wide range of experimental designs

Reduces sequencing costs and turnaround time compared to broader approaches such as whole-genome sequencing

Roche 454 pyrosequencing by synthesis (SBS) was the first commercially successful second generation sequencing system developed by 454 Life Sciences in 2005 and acquired by Roche in 2007 http://www.my454.com

This technology uses sequencing chemistry, whereby visible light is detected and measured after it is produced by an ATP sulfurylase, luciferase, DNA polymerase enzymatic system in proportion to the amount of pyrophosphate that is released during repeated nucleotide incorporation into the newly synthesized DNA chain

The estimated cost per million bases is $10 by Roche 454 compared to $0.07 by Illumina HiSeq 2000.

A more serious challenge for those using this technology is the announcement by Roche that they will no longer supply or service the 454 sequencing machines or the pyrosequencing reagents and chemicals after 2016

also sequencing by synthesis

• previous model Genome Analyzer can sequence 1Gb per run
• Solexa model can sequence 600 Gb per run
• HiSeq model 100 bases read length average 30Gb per run
  • 11 day in regular mode
  • 2 day in rapid run mode

and IMPROVING…

• ∼$1000 MiSeq
• ∼$3000 HighSeq
• ∼ 1.5 days

Illumina http://www.illumina.com purchased the Solexa Genome Analyzer in 2006 and commercialized it in 2007

Today, it is the most successful sequencing system with a claimed over 70% dominance of the market, particularly with the HiSeq and MiSeq platforms

The Illumina sequencer is different from the Roche 454 sequencer in that it adopted the technology of sequencing by synthesis using removable fluorescently labeled chain-terminating nucleotides that are able to produce a larger output at lower reagent cost

Supported Oligonucleotide Ligation and Detection (SOLiD) is a next-generation sequencer instrument marketed by Life Technologies http://www.lifetechnologies.com and first released in 2008 by Applied Biosystems Instruments (ABI)

It is based on 2-nucleotide sequencing by ligation (SBL)

This procedure involves sequential annealing of probes to the template and their subsequent ligation

Sequencers on the market today, such as the 5500 W series, are suitable for small- and large-scale projects involving whole genomes, exomes, and transcriptomes

Previously, sample preparation and amplification was similar to that of Roche 454 sequencing

However, the upgrades to Wildfire chemistry have enabled greater throughput and simpler workflows by replacing beads with direct in situ amplification on FlowChips and paired-end sequencing

The SOLiD 5500 W series sequencing reactions still use fluorescently labeled octamer probes in repeated cycles of annealing and ligation that are interrogated and eventually deciphered in a complex subtractive process using Exact CallChemistry that has been well described by others

The advantage of this method is accuracy with each base interrogated twice

The major disadvantages are the short read lengths (50–75 bp), the very long run times of 7 to 14 days, and the need for state-of-the-art computational infrastructure and expert computing personnel for analysis of the raw data

Complete Genomics http://www.completegenomics.com developed DNA nanoball sequencing (DNBS) as a hybrid of sequencing by hybridization and ligation

Small fragments (440–500 bp) of genomic DNA or cDNA are amplified into DNA nanoballs by rolling-circle replication that requires the construction of complete circular templates before the generation of nanoballs

The DNA nanoballs are deposited onto an arrayed flow cell, with one nanoball per well sequenced at high density

Up to 10 bases of the template are read in 5′ and 3′ direction from each adapter

Since only short sequences, adjacent to adapters, are read, this sequencing format resembles a multiplexed form of mate-pair sequencing similar to using Exact Call Chemistry in SOLiD sequencing

Ligated sequencing probes are removed, and a new pool of probes is added, specific for different interrogated positions

The cycle of annealing, ligation, washing, and image recording is repeated for all 10 positions adjacent to one terminus of one adapter

This process is repeated for all seven remaining adapter termini

Although the developers have sequenced the whole human genome, the major disadvantage of DNBS is the short length of reads and the length of time for the sequencing projects

Claimed cost of the reagents for sequencing of the whole human genome is under $5000

The major advantage of this approach is the high density of arrays and therefore the high number of DNBs (~350 million) that can be sequenced

In 2015, the Chinese genomics service company BGI-Shenzhen acquired Complete Genomics and introduced the Retrovolocity system for large-scale, high quality whole-genome and whole-exome sequencing with 50x coverage per genome and with the sample to assembled genome produced in less than 8 days

Complete Genomics claims to have sequenced more than 20,000 whole human genomes over 5 years and published widely on the use of their NGS platform

They provide public access to a human repository of 69 genomes data and a cancer data set of two matched tumor and normal sample pairs at http://www.completegenomics.com/public-data/

Ion Torrent technology http://www.iontorrent.com was developed by the inventors of 454 sequencing, introducing two major changes.

Firstly, the nucleotide sequences are detected electronically by changes in the pH of the surrounding solution proportional to the number of incorporated nucleotides rather than by the generation of light and detection using optical components.

Secondly, the sequencing reaction is performed within a microchip that is amalgamated with flow cells and electronic sensors at the bottom of each cell

The incorporated nucleotide is converted to an electronic signal detected by the electronic sensors

The two sequencers in the market that use Ion Torrent technology are the high-throughput Proton sequencer with more than 165 million sensors and the Ion Personal Genome Machine (PGM), a bench-top sequencer with 11.1 million sensors

There are four sequencing chips to choose from

The Ion PI Chip is used with the Proton sequencer, and the Ion 314, 316, or 318 Chips are used with the Ion PGM.

The Ion 314 Chip provides the lowest reads at 0.5 million reads per chip, whereas the Ion 318 Chip provides the highest reads of up to 5.5 million reads per chip

The Proton sequencer provides a higher throughput (10–100 Gb vs. 20 Mb–1 Gb) and more reads per run (660 Mb vs. 11 Mb) than the PGM chips, but the read lengths (200–500 bp), run time (4–5 h), and accuracy (99%) are similar

Sample preparation for the generation of DNA libraries is similar to the one used for Roche 454 sequencing but can be simplified with the use of the Ion Chef system for automated template preparation and chip loading

The Ion Torrent chip is used with an ion-sensitive field-effect transistor sensor that has been engineered to detect individual protons produced during the sequencing reaction

The chip is placed within the flow cell and is sequentially flushed with individual unlabeled dNTPs in the presence of the DNA polymerase

Incorporation of nucleotide into the DNA chain releases H protons and changes the pH of the surrounding solution that is proportional to the number of incorporated nucleotides

The major disadvantages of the system are problems in reading homopolymer stretches and repeats

The major advantages seem to be the relatively longer read lengths, flexible workflow, reduced turnaround time, and a cheaper price than those provided by the other platforms.

• PacBio model RS II produces average read lengths over 10 kb, with an N50 of more than 20 kb and maximum read lengths over 60 kb
• Error rate of a continuous long read is relatively high (around 11%–15%)
• The error rate can be reduced by generating circular consensus sequences reads with sufficient sequencing passes.

Faruk

• Jerzy K. Kulski Next-Generation Sequencing — An Overview of the History, Tools, and “Omic” Applications (2016) DOI: 10.5772/61964

Esra

• Glenn et al. Molecular Ecology Resources, 2011 May, doi.org/10.1111/j.1755-0998.2011.03024.x
• ChIPBase, School of Life Science, Sun Yat-sen University, China.
• ODIN, detecting differential peaks in ChIP-Seq. DOI: 10.1093/bioinformatics/btu722.

Esra

• Practical Guidelines for the Comprehensive Analysis of ChIP-seq Data, https://doi.org/10.1371/journal.pcbi.1003326
• Probabilistic Inference on Multiple Normalized Signal Profiles from Next Generation Sequencing: Transcription Factor Binding Sites, https://doi.org/10.1109/TCBB.2015.2424421
• SignalSpider: probabilistic pattern discovery on multiple normalized ChIP-Seq signal profiles, https://doi.org/10.1093/bioinformatics/btu604

• Always include references to your sources. This is essential in science
  • Not doing it is unethical
  • Doing it gives you credibility
• Summarize. Don’t make me think
  • What do the reader needs to know
  • Put yourself in the reader’s place
• So what? What are the consequences?
  • How can we use your data?
  • Information is the part of data that allows us to make a decision

• You want to sequence an organism. You know the approximate genome size
  • What technology will you use? Why?
  • How long will it take?
  • How many contigs do you expect?
  • What will be the average depth?
  • What will be the N50?
  • How much will it cost?

As we saw in the last class, mapping reads to a reference genome is a useful technique

• It can be use to assemble a closely related organism
• It can be use to detect polymorphisms

Mapping is finding where each read maps on the genome

Here “maps” means “the genome sequence is very similar to the read”

The idea is that

• if the genome is fragmented,
• and the read is part of the genome

then we want to know where the fragment came from

For now, let’s assume that the read is 100% identical to a part of the genome

The problem is identical to “finding a word in a text”

How does CTRL-F works?

The time of a program is usually called its computational cost

It depends on the genome length \(G\) and the read length \(L\)

There are two loops. The inner part of the loop is done \(G\cdot L\) times

We say that the cost of this program is \(O(GL)\)

Cost is “on the order of” \(G\cdot L\)

For example, for a genome of 3 million bp and a read of 500 bp, the cost is on the order of 1.5E9 steps

That is 1.500.000.000 steps for one read

We have to look for several reads. Millions. Let’s say \(N\)

Then we have to repeat everything \(N\) times

If we have 2 million reads, we multiply 1.5E9 by 2E6

That is 3E15 = 3.000.000.000.000.000 steps

Assuming that the computer can do 1000 million steps per second, this is \[3\cdot 10^{15}\text{ steps}\cdot \frac{1\text{ sec}}{10^{9}\text{ steps}}\cdot\frac{1\text{ day}}{86400\text{ sec}}\approx 34.7\text{ days}\]

Most of the algorithms used in bioinformatics were initially created for other purposes

Here we are using ideas that were initially developed for searching patterns (such as words) inside long texts, like in a book

In our context, the text will be the full genome, and the words are subsequences of DNA of a fixed length.

Usually, the word length is called k, and the words are called k-mers

To look for a read in the genome, we see which k-mers are in the read, and we look for them first

Searching in the index is easy, because it is sorted

When data is sorted, we can search quickly by dividing the list in half every step

This is called Binary Search. Google it

Using this strategy of dividing the index in half on every step, we need \[\log_2(\text{Number of words in the index})\cdot (\text{word length})\] steps to find any word in the index

It is easy to see that there are \(4^k\) different k-mers

Some may appear several times on the genome, some may not appear

On average, each k-mer appears \(G/4^k\) times on the genome

In any case, the number of words in the index is \(4^k\)

Using the strategy of binary search, we can find each word in the index

\[\log_2(4^k)\cdot k=\log_2(2^{2k})\cdot k=2k^2\]

After we find the k-mer in the index, we have to see it the rest of the read is on the genome

This time we do not need to see all the genome

we have to look only \(G/4^k\) positions on the genome

As before, testing to see if the read is on any position takes \(L\) steps

The total cost of a single read is then \[2k^2\cdot\frac{G}{4^k}\cdot L\]

As we can see, using an index is much faster. But we need to make and store the index

Can you estimate how many steps do we need to make the index?

Can you estimate how much memory is needed?

Obviously, the answer depends on the value of \(k\)