mjpdejong

Nature Methods 12, 303 (2015). doi:10.1038/nmeth.3327

Authors: Nicholas J Loman & Mick Watson

Nanopore sequencing gets a boost with accurate error modeling and variant-calling tools for Oxford Nanopore Technology's highly anticipated MinION platform.

An unusual and very exciting form of carbon -- that can be created by drawing on paper -- looks to hold the key to real-time, high throughput DNA sequencing, a technique that would revolutionize medical research and testing.

Messenger RNAs -- the working copies of genes that are used to assemble proteins -- have typically been ignored as drug targets because they all look about the same. But researchers have found that a subset of mRNAs -- many of which have been linked to cancer -- have unique tags. These short RNA tags bind to a protein, eIF3, that regulates translation at the ribosome, making the binding site a promising target for anticancer drugs.

Our AGBT workshop attracted more than 500 attendees thanks to the high-profile speakers who shared their perspectives on human genomic research. Because of the exclusivity of AGBT, we decided to live-stream our workshop to reach the broader scientific community. Thanks to the the hundreds of people who tuned in to our live webcast from afar! Here are some highlights from the presentations and the recording of the workshop is at the bottom of this post.



Our CEO, Mike Hunkapiller, started the session with a reflection on the 15-year anniversary of the announcements of the first human genomes, noting these efforts required considerable effort and produced draft assemblies with contig N50s in the 20-24 kb range. Many technologies and methods have been introduced since then, but assembly quality has not improved dramatically and scientists are still missing critical genomic information. He noted structural variations, in particular, have been underestimated, limiting our understanding of human genomes. He then unveiled a PacBio® diploid assembly of Craig Venter’s genome, chosen because it has been so well characterized over the years. Compared to the original iteration of the Venter assembly, the PacBio diploid assembly contains 3004 primary contigs, a contig N50 of 10.4 Mb, and the longest contig is 34.6 Mb. The 4,761 associated contigs, representing potential structural variants, total 189 Mb with a mean length of 39.8 kb.

In his first appearance at Marco Island, Venter (accompanied by his dog, Darwin) offered his vision for Human Longevity, Inc. (HLI), as well as at the J. Craig Venter Institute and Synthetic Genomics. At HLI, his team plans to sequence 1 million genomes in the coming years while also gathering extensive phenotypic information to make meaningful connections from the data. To support this effort, they will produce 30 reference genomes representing ethnogeographic diversity. Venter told attendees the PacBio machine gives you a great reference genome.

Next up was Gene Myers from the Max-Planck Institute, who addressed the concept of a near-perfect human assembly. He believes this level of quality is within reach, made possible by the long reads, random error, and random sampling of SMRT® Sequencing. Myers and his team have been working hard to build new analysis tools for processing this data, including a lightning-fast aligner and a scrubbing algorithm. His tools are available through his Dazzler website.

Deanna Church from Personalis spoke about the importance of a complete, truly representative human reference genome. Having this data is necessary for calling and interpreting variants, noting something as simple as a missing gene in the reference can confound other calls in a new reference-based assembly. She championed the new regions of alternative loci available with the GRCh38 human reference, saying this sequence is essential to ensure you’re not missing valuable information in genome interpretation. Church urged attendees to generate high-quality sequence assemblies and contribute them back to the databases to continue refinement of the reference genome.

Jeong-Sun Seo, CEO of Macrogen, Inc., spoke about his team’s efforts to sequence large numbers of Asian genomes and the generation of a representative diploid human genome reference for the Asian population. His team used PacBio technology, PacBio’s latest diploid assembly methods, optical mapping from BioNano, and BAC sequencing to create the most comprehensive genomic reference possible for a Korean human genome sample. He showed examples of gaps that could be closed or extended within the GRC38 reference thanks to SMRT Sequencing data, and highlighted work to identify structural variants, some of which are implicated in diseases that affect Asian populations more than other populations.

Finally, Dick McCombie from Cold Spring Harbor Laboratory presented work on a breast cancer cell line known to be riddled with rearrangements, amplifications, and other complex events. Working with collaborators at OICR, he is using long-read sequencing to generate a higher-resolution view of the structural variation occurring in this cell line. The project, which began last November and is still ongoing, has already led to promising results, such as detecting complex structural variants missed by short-read sequencers. A de novo assembly generated by DNANexus in 22 hours produced an unprecedented contig N50 of 2.56 Mb. Download the raw data from the Schatz lab website.


 

 OpenSource.com: We take a look at the biggest news from the Game Developers Conference

ZDnet: Linux software developers are working hard on expanding Linux's file and storage options.

The Raspberry Pi 2 is a lot faster than its predecessors , but you still might want to overclock it for the best possible performance. Blogger Hayden James breaks down what each overclock setting does and shares some configurations for better performance.

Read more...

Sjon shared this story from Hacker News 100.

---

Comments

MakeUseOf: Make ASCII art, talk to your computer and play text adventures.

Encoded libraries of chemically modified peptides.

Curr Opin Chem Biol. 2015 Mar 10;26:89-98

Authors: Heinis C, Winter G

Abstract
The use of powerful technologies for generating and screening DNA-encoded protein libraries has helped drive the development of proteins as pharmaceutical ligands. However the development of peptides as pharmaceutical ligands has been more limited. Although encoded peptide libraries are typically several orders of magnitude larger than classical chemical libraries, can be more readily screened, and can give rise to higher affinity ligands, their use as pharmaceutical ligands is limited by their intrinsic properties. Two of the intrinsic limitations include the rotational flexibility of the peptide backbone and the limited number (20) of natural amino acids. However these limitations can be overcome by use of chemical modification. For example, the libraries can be modified to introduce topological constraints such as cyclization linkers, or to introduce new chemical entities such as small molecule ligands, fluorophores and photo-switchable compounds. This article reviews the chemistry involved, the properties of the peptide ligands, and the new opportunities offered by chemical modification of DNA-encoded peptide libraries.

PMID: 25768886 [PubMed - as supplied by publisher]

Wired: Chris DiBona worried everything would end up in one place.

A new technique for creating artificial DNA that is faster, more accurate and more flexible than existing methods has been developed by scientists.

Researchers have extracted and sequenced tiny bits of DNA remaining in the teeth of 300-year-old skeletons in the Caribbean. From this data, they were able to determine where in Africa the individuals likely lived before they were captured and enslaved.

Sjon shared this story from Hacker News 100.

---

Comments

---

Click here for all of GameBeat’s 2015 Game Developers Conference coverage. 

Ubisoft, the maker of games like Assassin’s Creed, and healthtech startup Amblyotech have teamed up to create Dig Rush, a therapeutic video game that can help treat a condition known as “lazy eye.” With “lazy eye,” patients see out of one eye instead of two.

The title from the French game publisher is part of a trend of “games for health,” or those that can be used to treat health problems, and it is being unveiled today at the Game Developers Conference in San Francisco. Its creators say that the treatment is 90 percent effective, and it takes only four to six weeks to retrain your brain to use both eyes instead of one eye, based on clinical testing.

Amblyopia, or “lazy eye,” is where one eye dominates a person’s vision to the point where they don’t have much depth perception and see life effectively through a single eye. The condition affects 3 percent of the population, or about 9 million people in the U.S.

“You know us for developing triple-A games, but we’re also producing engaging games for other purposes,” said Mathieu Ferland, the senior producer at Uibisoft’s studio in Montreal, Canada. “This game is a good demonstration that the positive impact video game technology can have on our society.”

Dig Rush was developed with researchers at McGill University in Montreal, and with partner Amblyotech in Atlanta, Ga. The latter company is in charge of applications to the Food and Drug Administration for federal approval of the game as a treatment.

I saw a demo of the game, which is patented. It uses 3D glasses and an app on an Android or iOS tablet. The game’s interactive objects are shaded as either red or blue objects. But the 3D glasses allow you to see the red objects only with your right eye. You can see the blue objects with your left eye. I have lazy eye. So when I first put on the glasses, I couldn’t see the blue objects. When I concentrated, I used my left eye and could see the blue objects along with the red ones.

“The only way to play the game is to force the patient to use binocular vision, or two eyes,” Ferland told GamesBeat.

Robert Derricotte, the chief operating officer of Amblyotech, said the companies are seeking FDA clearance now.

“With this condition, the brain suppresses everything from one eye, and you have no depth of field perception,” he said. “We fix that, using both eyes in order to be able to regain the images coming out of the suppressed eye. Your 3D acuity is regained.”

He said the treatment is effective compared to patching, where a patient covers a stronger eye up with a patch for six months to four years. Patching doesn’t work on older patients, and it doesn’t retrain the brain as well as the video game can, Derricotte said.

“We’re very excited about this,” Derricotte said. “This is a game changer. Doctors have been patching patients for over 200 years.”

Dig Rush uses inventions patented by McGill University researchers Robert Hess, Benjamin Thompson, Behzad Mansouri, Jeremy Cooperstock, Long To, and Jeff Blum. They licensed the technology to Amblyotech, and Ubisoft is applying its skills with gameplay to create an entertaining way to implement the treatment.

Ubisoft has been working on the title for a year. Joseph Koziak, the CEO of Amblyotech, said in an interview, “Quite frankly, this is a drug. We are making that statement. And if you say that, you have to get FDA approval. It’s not just a game. The game part is for your attention. The therapy is the contrast levels that occur throughout the hours of play. It changes the balance so that you no longer have a dominant eye.”

“We think we can create a hybrid industry between gaming and the life sciences,” said Koziak. “The tablet is the syringe of the future. You can use an iPad visual display to administer a drug to a patient. Other medical conditions of the future can be treated through a display.”

Other researchers have been looking at the problem. Diplopia is working on a demo for treating “lazy eye” through Oculus Rift virtual reality goggles. Koziak and Derricotte are trying to create a whole company around novel medical treatments using smartphones and tablets.

“McGill University has a proud history of innovation and product development in numerous fields, especially life sciences,” said Michèle Beaulieu, the associate director of the Invention Development and Entrepreneurship Assistance Team in the Office of Innovation and Partnership of McGill University, in a statement. “It is our pleasure to see one of our invented technologies take the next step to commercialization.”

Scout: Your Linux server is running slow, so you follow standard procedure and run top.

  During the AGBT meeting this February at Marco Island I decided to “get modern” and joined Twitter in part due to the encouragement of my younger colleagues, Ned Lederbragt and Pavel Tomancak.  I was at the AGBT meeting to present my new work towards an assembler for long-read-only data sets.  I’ve been out of the DNA sequencing “scene” for almost ten years now, mostly because the trend has been so much on cost versus quality that I just lost interest as a mathematician.  With the short read instruments, a great assembly is just not possible, especially in the face of strong sampling bias of the reads and consistently reproducible sequencing errors.  Yes, the results are still biologically useful, but I’m not a biologist.  What mathematician wants to work on a problem they know they cannot solve to their satisfaction?  Moreover, as a bioinformatician I know how much time is spent (wasted) dealing with the overwhelming number of caveats in the data sets.  Wouldn’t it be great if at least the source of much our work, the DNA sequence, was perfect?

What I perceived early in 2013 was that the relatively new Pacbio “long read” DNA sequencer was reaching sufficient maturity that it could produce data sets that might make this possible, or at least get us much, much closer to the goal of near perfect, reference quality reconstructions of novel genomes.  While the accuracy of its reads is relatively low (85-88%), causing many to erroneously think it not a good machine, the Pacbio machine has two incredibly powerful offsetting properties, namely, that (i) the set of reads produced is a nearly Poisson sampling of the underlying genome, and (ii) the location of errors within reads is truly randomly distributed.   Given my previous experience, I immediately understood the implication of these two properties, and decided in the summer of 2013 to buy a machine and get back into the genome assembly game.

So with my new twitter account, I decided to float the following 140 character theorem, that encapsulates my reasoning for coming back to the DNA sequencing scene:

Thm: Perfect assembly possible iff a) errors random b) sampling is Poisson c) reads long enough 2 solve repeats. Note: e-rate not needed

There was much follow on: some people didn’t believe it and some were (quite reasonably) confused about some of the conditions (especially c).  In addition I got some interesting links to work along these lines (here and here).  So how did I get to the conclusion of the theorem?

Condition (b), by the Poisson theory of Lander and Waterman, implies that for any minimum target coverage level k, there exists a level of sequencing coverage c that guarantees that every region of the underlying genome is covered k times.  In their 1988 paper they showed that the amount of the genome uncovered was roughly e^-c where e is Napier’s constant, and an easy generalization shows that the amount of the genome not covered at least k times is c^k/k! e^-c.  Certain as c goes to infinity this quantity goes to 0 and since the genome is finite, at some point every part of the genome is covered k times.

Condition (a), from the early work of Churchill and Waterman, implies that the accuracy of the consensus sequence of k sequences, each with average error err, is O(err^k) which goes to 0 as k increases.  This clearly implies that one can arrive at an arbitrarily accurate consensus by sampling enough sequences.  Note carefully that this is not true for a device that has reproducible errors and that every previous sequencing device makes such reproducible errors, the best ones at about a rate of 1 per 10Kbp.  This implies that with previous devices the best accuracy you could ever expect is a “Q40” reconstruction.  For the Pacbio where error is truly random any desired accuracy is possible.  For example, the Pacbio team has produced a Q60 reconstruction of E. Coli with their machines.

So conditions (a) and (b) together imply that with sufficiently large coverage c, an arbitrarily accurate reconstruction of the entire genome is possible, provided the sequencing reads can be correctly assembled.  As is well known it is the underlying repeat structure of a genome that confounds assembly.  Condition (c) is very loosely stated and basically says, if the reads are so long that every repetitive region is spanned, then the repeats aren’t a problem.  Of course, one could try to characterize more carefully how long the reads need to be (for example, they tried here), or one could simply capitulate and say all regions not involving repeats over a given size and fidelity are correctly assembled.  Practically, full length retrotransposons, the most frequent interspersed repetitive element, are 7Kbp long so most regions of any genome will not be problematic with 10Kbp reads.  Typically one finds that the euchromatic arms of most genomes assemble correctly into one or perhaps a few pieces with reads of this length.  While one can try to make theoretical statements, the truth of the matter is most assemblers are still not good at resolving repeats on the basis of the micro-variation within them, i.e. there is still significant room for improvement in the combinatorial aspects of genome assembly.  We intend to aggressively address this weakness going forward.

The theorem presupposes that there is a single haplotype under consideration.  Diploidy and polyploidy create additional complications that I believe are addressable.  But the really important issue is how to build an assembler that is efficient in coverage c.  That is, the theorem just says for c “big” enough.  But pragmatically the smaller c, the better the cost and efficiency.  Currently most of the “Pacbio-only” projects have involved 50X or more.  But in principle, a 20X assembly should be pretty darn good.  But no one currently gets good assembly at 20X because the 50X is needed to “correct” reads, i.e. understand read error, sufficiently to put the pieces together.  So there is lots of room for innovation and improvement in the arena of genome assembly algorithms: assembly is not a solved problem !

Sure it sounds like something from the book Frankenstein, but Sergio Canavero of the Turin Advanced Neuromodulation Group has made it known that he intends to announce at this summer’s American Academy of Neurological and Orthopedic Surgeons meeting, that he believes he has put together a group of techniques that should make it possible to […]...

Posted by Patrick Riley and Dale Webster, Google Research and Bharath Ramsundar, Google Research Intern and Stanford Ph.D. candidate

Discovering new treatments for human diseases is an immensely complicated challenge; Even after extensive research to develop a biological understanding of a disease, an effective therapeutic that can improve the quality of life must still be found. This process often takes years of research, requiring the creation and testing of millions of drug-like compounds in an effort to find a just a few viable drug treatment candidates. These high-throughput screens are often automated in sophisticated labs and are expensive to perform.

Recently, deep learning with neural networks has been applied in virtual drug screening^1,2,3, which attempts to replace or augment the high-throughput screening process with the use of computational methods in order to improve its speed and success rate.⁴ Traditionally, virtual drug screening has used only the experimental data from the particular disease being studied. However, as the volume of experimental drug screening data across many diseases continues to grow, several research groups have demonstrated that data from multiple diseases can be leveraged with multitask neural networks to improve the virtual screening effectiveness.

In collaboration with the Pande Lab at Stanford University, we’ve released a paper titled "Massively Multitask Networks for Drug Discovery", investigating how data from a variety of sources can be used to improve the accuracy of determining which chemical compounds would be effective drug treatments for a variety of diseases. In particular, we carefully quantified how the amount and diversity of screening data from a variety of diseases with very different biological processes can be used to improve the virtual drug screening predictions.

Using our large-scale neural network training system, we trained at a scale 18x larger than previous work with a total of 37.8M data points across more than 200 distinct biological processes. Because of our large scale, we were able to carefully probe the sensitivity of these models to a variety of changes in model structure and input data. In the paper, we examine not just the performance of the model but why it performs well and what we can expect for similar models in the future. The data in the paper represents more than 50M total CPU hours.

This graph shows a measure of prediction accuracy (ROC AUC is the area under the receiver operating characteristic curve) for virtual screening on a fixed set of 10 biological processes as more datasets are added.

One encouraging conclusion from this work is that our models are able to utilize data from many different experiments to increase prediction accuracy across many diseases. To our knowledge, this is the first time the effect of adding additional data has been quantified in this domain, and our results suggest that even more data could improve performance even further.

Machine learning at scale has significant potential to accelerate drug discovery and improve human health. We look forward to continued improvement in virtual drug screening and its increasing impact in the discovery process for future drugs.

Thank you to our other collaborators David Konerding (Google), Steven Kearnes (Stanford), and Vijay Pande (Stanford).

References:

1. Thomas Unterthiner, Andreas Mayr, Günter Klambauer, Marvin Steijaert, Jörg Kurt Wegner, Hugo Ceulemans, Sepp Hochreiter. Deep Learning as an Opportunity in Virtual Screening. Deep Learning and Representation Learning Workshop: NIPS 2014

2. Dahl, George E, Jaitly, Navdeep, and Salakhutdinov, Ruslan. Multi-task neural networks for QSAR predictions. arXiv preprint arXiv:1406.1231, 2014.

3. Ma, Junshui, Sheridan, Robert P, Liaw, Andy, Dahl, George, and Svetnik, Vladimir. Deep neural nets as a method for quantitative structure-activity relationships. Journal of Chemical Information and Modeling, 2015.

4. Peter Ripphausen, Britta Nisius, Lisa Peltason, and Jürgen Bajorath. Quo Vadis, Virtual Screening? A Comprehensive Survey of Prospective Applications. Journal of Medicinal Chemistry 2010 53 (24), 8461-8467

Investigators administered a customized genetic construct consisting of tiny rings of DNA, called DNA minicircles, to mice. The scientists then showed that mice with tumors produced a substance that tumor-free mice didn't make. The substance was easily detected 48 hours later by a simple blood test.

A study that examined 17 million mutations in the genomes of 650 cancer patients concludes that large differences in mutation rates across the human genome are caused by the DNA repair machinery. 'DNA spellchecker' is preferentially directed towards more important parts of chromosomes that contain key genes. The study illustrates how data from medical sequencing projects can answer basic questions about how cells work.

Current approaches in SELEX: An update to aptamer selection technology.

Biotechnol Adv. 2015 Feb 20;

Authors: Darmostuk M, Rimpelová S, Gbelcová H, Ruml T

Abstract
Systematic Evolution of Ligands by Exponential Enrichment (SELEX) is a well-established and efficient technology for the generation of oligonucleotides with a high target affinity. These SELEX-derived single stranded DNA and RNA molecules, called aptamers, were selected against various targets, such as proteins, cells, microorganisms, chemical compounds etc. They have a great potential in the use as novel antibodies, in cancer theragnostics and in biomedical research. Vast interest in aptamers stimulated continuous development of SELEX, which underwent numerous modifications since its first application in 1990. Novel modifications made the selection process more efficient, cost-effective and significantly less time-consuming. This article brings a comprehensive and up-to-date review of recent advances in SELEX methods and pinpoints advantages, main obstacles and limitations. The post-SELEX strategies and examples of application are also briefly outlined in this review.

PMID: 25708387 [PubMed - as supplied by publisher]

Researchers have long sought an efficient way to untangle DNA to study its structure -- neatly unraveled and straightened out -- under a microscope. Now, researchers have devised a simple and effective solution: they inject genetic material into a droplet of water and use a pipet tip to drag it over a glass plate covered with a sticky polymer.

The Financial Times has taken a look inside Chinese sequencing giant BGI and the implications of its mission to sequence the genomes of one million humans, one million microorganisms and one million plants and animals. 

How can we preserve our knowledge today for the next millennia? Researchers have found a way to store information in the form of DNA, presumably preserving it for nearly an eternity.

Pictures courtesy of Lucasfilm and J. Armbruster     Leaving the movie theater in 1977, with Greedo's death at the hands of Han Solo a fresh memory, a young Jon Armbruster could not have anticipated the role that Jabba the Hutt's go-to bounty hunter would play in his scientific contributions decades later.     And yet...when he (along with Auburn University researchers Milton Tan, Christopher...

Related Articles

Single-step selection of bivalent aptamers validated by comparison with SELEX using high-throughput sequencing.

PLoS One. 2014;9(6):e100572

Authors: Wilson R, Bourne C, Chaudhuri RR, Gregory R, Kenny J, Cossins A

Abstract
The identification of nucleic acid aptamers would be advanced if they could be obtained after fewer rounds of selection and amplification. In this paper the identification of bivalent aptamers for thrombin by SELEX and single-step selection are compared using next generation sequencing and motif finding informatics. Results show that similar aptamers are identified by both methods. This is significant because it shows that next generation sequencing and motif finding informatics have the potential to simplify the selection of aptamers by avoiding multiple rounds of enzymatic transcription and amplification.

PMID: 24963654 [PubMed - indexed for MEDLINE]

Borrowing the method organisms use to build mineralized tissues could lead to better performing lithium ion batteries

-- Read more on ScientificAmerican.com

Researchers have designed a tiny cage that can trap a single strand of DNA after it has been pulled through a nanopore. While caged, biochemical experiments can be performed on the strand, which can then be zipped back through the nanopore. The device could enable researchers to probe DNA before and after a reaction takes place.

Scientists have created a new research tool, based on the fruit fly, to help crack the histone code. This research tool can be used to better understand the function of histone proteins, which play critical roles in the regulation of gene expression in animals and plants.