It feels like many of us were so excited about the automation scenarios, we may have missed what the significance of the latest developments in machine-learning (both algorithmic and hardware-based) would have in other fields - or maybe it's just me.
Article starts with discussing how recent improvements in computing led to the start of the expected discoveries in genetics. I don't understand how my comment is unrelated?
I changed "path to singularity" to "automation scenarios" and "IT people and target audience of HN" to "many of us" to make it more clear. I'm very sorry if you think it meant something else with those words, I'm not a native English speaker.
What are "automation scenarios"? Saying people on HN discuss the singularity too often vs "automation scenarios" is going from a specific idea to a vague idea. English is my first language and I would never say "automation scenarios"
Programming is mainly about automation and I mean "automation scenarios" as the general concepts we work on. Improving automation is a path towards "singularity" (in this case that meaning not AI world domination, but nothing to do left for humanity as everything is being done automatically) with a nice feedback loop. Many people here, including me, naturally obsess over the parts that are directly relevant to our jobs and not think too much about the possible effect on other industries. Hope it is clear now :)
The costs of assaying a genome has dropped from $100 million to $1 thousand.
Identifying correlations with clinical metadata, understanding the post translational interactions and understanding the full roles of proteins and other molecules remains an intimidating and difficult job.
The $1000 is also minor compared to the computational and personnel costs required to interpret a clinical genome. It's almost cheaper to re-sequence a genome than store the raw data long term. And the interpretation of the genome usually requires a multi-disciplinary team with advanced degrees (MS/PhD/MD). That part of the problem is still largely a manual process, and will be for the foreseeable future.
I'm not saying that it's not worth it in some cases, but the sequencing is only a small part of the overall costs.
But at least scaling computation / storage is possible. A challenge to be sure, but it's a challenge that the entire computer industry is dealing with.
Interpretation, on the other hand, is quite different. Currently, interpretation still requires a human to look at everything. And regardless of how good annotations and algorithms get, I have yet to hear of anyplace where the final annotations aren't manually curated (maybe automated, but still requires a human in the loop). This aspect of Precision Medicine doesn't scale quite as well as sequencing or computation...
I kind of hate articles like this. They feed into the myth that single event cause revolutionary changes in the way the world works, when it reality its thousands of evolutionary events that cause change.
I'm not trying to belittle recent advances in genomics, on the contrary I believe that cheap sequencing and technologies like CRISPR are going to change the world. However, its the insinuation that the last 20+ years of hard work were a fraud that upsets me. We wouldn't be were we are now without the last 20 years of genomics research. /rant
The article kind of says that though, the very beginning of the article explains how it was assumed that after we sequenced a genome, we'd make fast progress. However what's actually starting to make a difference is the sequencing of greater number of genomes (more data), and the improvement of our filtering (identification and algorithms). Basically, the big difference is a lot of hardwork, and refinement. There's not actually any mention of a single breakthrough discovery.
I think I get where the rant is coming from though. Even though the article isn't talking about a single discovery, it's more the idea that the world will turn one day and it will suddenly be the age of
Actually, cells have a lot of state. See [1] for a basic, slightly old, CS-biased explanation.
That's how a single genome yields an incredibly varied repertoire of cell types. We need to work a lot on understanding those epigenomes to then get a chance to pinpoint genetic causes of disease.
Yes, cell state is really important. My view is that nobody is looking at the big picture of state - cells store state in many different ways, with time periods of milliseconds to years. It's probably my CS background, but I think figuring out all the relevant state of a cell is important and neglected. Epigenetics is a part of this, but not enough.
RNAseq, the measure of abundance of transcribed genes at a particular cell state is one of the most common assays done with next gen sequencing. That's the big one then of course there is also miRs, transcription factors, methylation, histone marks, chromatin conformation capture.
As someone with a CS background who has worked almost 5 years in bioinformatics, one thing to realise is if you think something obvious is not done, either you're missing the right search term on pubmed or there's some horrible lab issue that makes it really hard (and we need a new assay/tech)
Exactly. Imagine being a student given a tome of knowledge called something like, 'Principia Mathematica'. Upon receipt of the book you say, "I will soon be able to make engines efficient, structures solid, understand the planet's rotation, and make mathematical models precise." Then two days later someone asks how to draw up plans for a rocket ship to the moon. You'll get there, but you have to read the text first, then you have to understand it, then you have to experiment with it, and then you will (go to the moon). Nothing in your prediction was unrealistic, but it takes time to learn.
A few decades ago ago we read our first line of source code from a computer from the year 2,000,000,000 - some pretty advanced technology that was just handed to us. Two decades ago we did our very first read-through of an entire human. The reading itself was not particularly productive. We already knew some of the most important lines of code, from the decades before, and we annotated a lot more sections that are critical for basic function. We've even traced a few bugs back to the source code - especially those bugs where 1) the source was unusually easy to read [1], or 2) the bug was really bad [2]. As of this year we've developed compilers to enable us to write and run small, multi-line programs of our own composition [3]. We're doing really well in trying to read, comprehend, manipulate, and act on some of the most sophisticated technology ever presented to us. More sophisticated by orders of magnitude than the silicon in Nvidia's new graphics card (16nm 2d dry feature size [4], vs 0.1nm, 3D wet feature size [5]) . Reverse engineering it takes a bit of time, and yet we've already put that knowledge to significant use.
Further, once we do get our working knowledge up and running there will be a significant outpouring of function from having reverse-engineered such technology. But of course there is a lag time in the millions of lab-hours being poured into understanding that technology before we can actually use it. Not dissimilar to the rapid advances in science-fiction when a race reverse-engineers an advanced alien technology. How long did it take us to build The Machine in Sagan's 'Contact'? We will not just cure cancer, but we will be capable of all sorts of new feats. Purchasing the textbook doesn't make one a master - it's the study, the hard work, and the practice.
The idea of lone geniuses single-handedly bringing upon revolutions in science and tech doesn't help, and it's perpetuated by our culture. Nobody thinks of the leagues of engineers or doctoral students or fellow "no-name" professionals who also do a lot of the work. It's really sort of fascinating how we idolatrize singular people who themselves will admit they were standing on the shoulders of giants.
I agree but only to a limited extent. There is such a thing as genius, it is very rare, and it is talent x infinity, which means no matter how many brilliant people you have in a room, they cannot match the insights and actions of a genius.
If the following had never lived, human history would not have progressed as it has: Galileo, Newton, Darwin (and Alfred Russell Wallace, I hasten to add), Einstein, Michael Polayni.
For anyone curious the idea that we make slow, discrete progress and that the idea that a few brilliant souls are responsible for most of the advances in human hostory is a bunch of hogwash, Matt Ridley has written an excellent book on the subject, The Evolution Of Everything (amazon.com/Evolution-Everything-How-Ideas-Emerge/dp/0062296000). Coincidentally he's also written a fabulous book on the human genome, either way it's great stuff and absolutely critical to understanding how to we can (as a species) progress most effectively.
Man, two Freudian slips before coffeetime... I need to better proof my posts. I even went to the wikipedia page to make sure I remember the name of 'The Machine' correctly.
Thanks for the rant! I'm with you. Many people probably would not read an article about incremental advances in science. As an economics topic, genomics may be turning into a bigger influence.
For a scientist working incrementally, accumulating data, sieving it for information, fighting for grants, doing administration while making sure family gets three squares, articles like this may cause a wince. But they probably influence support for more funding. Pro scientists I know do not have time to read these articles and will never see them unless it's from someone else, but budding scientists, career-changers, influencers, all get misinformed.
In short, news on science causes long-term damage for short-term gains.
To be fair, there has to some point in time you choose to claim that the "future has arrived". But I agree that great care should be taken to give credit where credit is due; and that means giving credit to all the failures as well, without which successes could never have arisen.
Thank you for your post. The best talk on CRISPR that I ever heard was ~45 minutes about previous genome editing techniques -- TALENs, Zinc finger proteins, etc -- and ~15 minutes about how CRISPR simplifies the genome editing process. It very much paid tribute to past work while also showing the utility of recent developments (while also noting that CRISPR is useless without extensive knowledge from genomics).
Yes. I'll start a thread to acknowledge the Bioinformatics programmer heroes of genomics, people who write the workhorse programs that biologists use everyday to assemble and analyze the data to come up with the biological results)
BLAST (Stephen Altschul, Warren Gish, Webb Miller, Eugene Myers, and David J. Lipman)
BWA (Heng Li, Richard Durbin)
Samtools (Heng Li)
GATK (from the GATK Team; DePristo, M., Banks, E., Poplin, R., Garimella, K., Maguire, J., Hartl., C., Philippakis, A., del Angel, G., Rivas, M.A, Hanna, M., McKenna, A., Fennell, T. Kernytsky, A., Sivachenko, A, Cibulskis, K., Gabriel, S., Altshuler, D. and Daly, M. A)
I'm missing quite a few from my more niche field; but feel free to add more.
I'd like to see the cost of bioinformatics cloud computing come down by several orders of magnitude as well. The ideal scenario seems to be a global network of in-field mobile clinical units that can collect samples, run diagnostic tests and sequence genome data. Then upload that collected real time data to public data stores on AWS S3 or Google Cloud Genomics. Which a team of researchers can then access from their own labs around the world.
Genomic data needs very strong privacy controls. HIPAA would probably prohibit making entire genomes public without patient consent, even if they have been de-identified. Just sequencing a fraction of the patient's genome subsequently could make identifying the entire public genome easy.
The About page says, "The 1000 Genomes Project developed guidelines on ethical considerations for investigators doing sampling, outlined in the Informed Consent Background Document and the Informed Consent Form Template. All collections included in the Project followed these ethical guidelines and model informed consent language."
The post I responded to sounded like the genomic data would be uploaded without consent to a giant public database.
Even with consent the greater concern is that employers, medical providers, national health services, and health insurance companies could, by sequencing part of your genome, match it to a public database identifying you and your entire genome. This could then be used to raise prices, deny health care, or deny employment.
De-identification is not enough if the material is indicative enough that efforts to re-identify it can succeed.
Kudos to those advancing science by making their genomes public, but it's a risk I would not take.
I'm not sure if the pendulum can swing any more in the direction of centralization. I expect genomic appliances that are disconnected from the internet to become common with advances in sequencing tech like nanopore sequencing. Imagine a lab that buys disposable sequencing based diagnostics. Open the package, drop in the sample, and get the answer anywhere and anytime without the need for a HiFi internet connection or HIPPA compliance woes.
The cost of bioinformatics cloud computing isn't going to come down by several orders of magnitude: it already tracks the raw hardware costs + cost of running a cloud data center + a tiny margin on top.
It's laughable that they led with the genetic cause for schizophrenia which (a) probably doesn't exist and (b) wasn't found a few months ago; they found a weak statistical association across a large number of variants that explains < 3.5% of the variability in disease susceptibility. What this demonstrates is what we've learned over and over across the past decade with association studies: yes, we're a product of our genes, and variation affects us, but no, common diseases are probably not mostly driven by genetics.
For me, the big story in genomics over the past few years is one of massive failure. We spent a lot of time and effort sequencing tumors in the hopes that the genetics would tell us something interesting and lead to cures. It did not. We now have a relatively complete cabinet of the major variants that drive tumors. Most of these variants we already knew about before we did these genomics studies (through older sequencing methods from decades prior), and most of what we learned tells us nothing new.
Genetics is at this point mostly garbage information. Why? Because we don't know what it means. We still don't understand gene expression, we don't understand signal transduction, and we can't understand the effects of mutations without painstaking characterization. All of these things mean our sudden wealth of knowledge of genetic variation tells us fuck-all about biology.
We learned quickly that 'cancer' is to the twentieth century 'the fevers'. We learned there are many reasons and causes for the catch-all, 'cancer'. Some we now know how to cure effectively. Some we now now how to prevent, effectively. Some we now know what to do to cure, but don't have good tools yet. And many we still do not understand. That is a significant improvement in a very short time span.
But there's a LOT more to learn than 'how to cure cancer'. We very clearly understand gene expression and signal transaction to a first order, and have bits of the second order down too. There might be more - but that doesn't make that first order inneffective or wrong. There's lots you can do with a first order understanding - see what Newtonian physics did for the world - it was correct only to the first order. The idea that we know 'fuck-all' about biology because of dna sequencing is just ridiculous. The ability to read and write dna is fueling one of the fastest growing segments of human progress today.
We didn't learn any of those things because of genomics. The major cancer therapies of today were developed before short read sequencing was even a thing; cancer researchers been saying for decades that cancer is "not one disease". In fact the genomics efforts were largely intended to overturn that wisdom by using genetic lesions to unify disparate cancer types (to little avail). About the only really interesting thing we learned from these recent efforts is that tumors are a lot more genetically heterogeneous than we previously appreciated, which just expands the problem space.
The proof in the pudding is this: you can get your exome sequenced today and report hundreds of tumor variants, but there is a bare handful of variants that will result in a change in your treatment, and most of those lesions were well-studied before genomics took off.
>We very clearly understand gene expression and signal transaction to a first order, and have bits of the second order down too. There might be more - but that doesn't make that first order inneffective or wrong.
Yes, in fact, our first order understanding IS ineffective. This is what I study every day, so perhaps I'm too close to this, but we literally do not understand the effect of most (99.9%) of genetic variants on gene expression. Sure, there's lots you can do with a first order understanding, but materially, what happens when you sequence a tumor (or a germline, for that matter), is that researchers get a list of mutations, stare at it mystified for a while, and then shrug and move on, because there's really nothing you can do to understand what these things mean.
Yes, this stuff is the way of the future, and it's important to do it for our greater knowledge and understanding in the future. But we're far, far away from this point right now.
I think your point is accurate, but jaded. Much as the above posters point out, there was no sudden 'click' where an individual discovered something and made genomics possible. And to the extent that's what we mean by 'genome sequencing', then you are correct - the shotgun approach coupled with big-data hasn't been very useful yet. However, if the message is more one of genetics at large, and how the ability to read/write DNA has affected our research, our discoveries, our therapeutics, and will affect our lives in the near-future, then your tone seems entirely too myopic. Nearly every drug you see on TV commercials today is a monoclonal antibody - which cannot have been made but for the ability to read and write DNA. Sure, it doesn't require the ability to read an entire animals' DNA at once (do the full genome sequences yet actually read every base in your genome even?). But it is a very real and very finite milestone on that exact same path. You are correct, genomic ≠ genetic, but in so far as they are similar (mostly), our understanding of genetics really has been a powerful driver in the past two to three decades.
Of course I recognize the potential of this technology; I wouldn't be in this field otherwise. And I probably am jaded, because I think people are constantly overemphasizing and overplaying our understanding of things in order to sound sexy and attract funding.
I think there's probably a fifteen-year lag in this stuff being useful. The insights we're having now are because of the sequencing of the human genome; the work we're doing now will pay off in another decade and a half when we've figured out how to deal with gene expression.
>do the full genome sequences yet actually read every base in your genome even?
This hysteria exist mainly because most people are led to believes "genes" are something that they are not.
They are not "blueprints" much less some sort of engineering document.
More realistically, they are a simple list of materials (in this case a listing of proteins).
They only describe the materials that make up a structure, not what that structure is nor how it functions.
Its the equivalent of someone saying: 14,234 tons of steel, 23,000 tons of concrete, 8000 tons of glass...etc. Now, what does it make? Why does it make it?
While this analogy is most likely not understood by people who don't understand gene expression, I still like it.
I think of genes as the initial configuration in Conway's game of life. Small initial (genetic) variations can cause large differences after some generations. Some don't matter at all and just die out. But no matter what is the end result, it's not a blue print or a description, it just _is_, and the result flows from it using some basic rules.
Epigenetics may influence gene expression, but as your link says others are skeptical that it constitutes an independent information vector. Epigenetic activities may themselves be encoded in genes.
Epigenetics as a concept can also be abused into a sort of neo-Lamarckism.
Physics. More specifically, 'conditions'. More specifically, the environment in which the ingredients are produced determines how the ingredients assemble. The analogy to the ingredients list is a decent one, but it is also a tad misleading. You generally do require knowledge of the environment to figure out how the parts will assemble, but many of the parts do work in many different environments and can be considered whole and separable objects rather than ingredients. But trying to separate evolution/life from its environment is the first mistake a student makes when studying evolution. As a grandparent stated, if the genes are the starting positions in Conway's Game of Life, the laws of physics are the rules of the game, and the physical conditions (temperature, pH, gravity, time, etc.) are the topology and specifics of the board you'll be playing on. Then you just hit play and certain concepts self-assemble - the flier, the block, the boat, etc. Some can then be used as objects in themselves, some can be used structurally, some can be used to build/destroy structure - that is life.
If the laws of physics are universal (fingers crossed), then this factor gets canceled out. Obviously organisms which evolved under our physical regime can only function properly within our physical regime. But this doesn't add any information as to why one set of genes manifests itself differently than another set of genes, since the physical laws are the same.
"Genes" as protein coding regions is a common shorthand, but there's another, broader definition of "gene" as "a locus in the genome that bears some responsibility for a trait". For instance, the ability to consume lactose post infancy is a trait driven by a mutation not in a protein coding region, but could reasonably called a "gene for lactase persistence". In general, we don't have good bottom-up ways of understanding which regions of the genome are regulatory, much less what that regulation is doing.
So according to one definition, it's even worse than your 14,234 tons of steel example: it's closer to "steel is a mixture of iron and carbon and trace amounts of other dopants, glass is silicon dioxide, concrete is ..." Without even listing the amounts. But by another definition, the genes are, if not a blueprint, at the very least a recipe and ingredient list. "Butter, flour, water, sugar, apples, cinnamon, cloves. Cut 125g butter into flour; if too coarse, continue cutting. Add water to ice. Add water from ice into flour..."
The information is clearly there—we have a long, unbroken history of offspring being very similar to their parents, despite having started with only a single, undifferentiated cell. But with the new tools available to us, we're starting to have a hope of understanding it, even if we're using more brute force than we'd like.
> More realistically, they are a simple list of materials (in this case a listing of proteins).
> Its the equivalent of someone saying: 14,234 tons of steel..
I feel like that's taking things a bit too far in the other direction. With our current understanding, it's like saying:
14,234 tons of steel that's capable of self-assembling (e.g., collagen), 23,000 tons of concrete that automatically folds into a specific structure (~everything other than disordered proteins).
That's not to say that we fully understand how proteins fold/function... but it's a bit more than a static block of steel.
I'm a bit surprised by your reaction. You do know we use genomics in clinical labs all the time, right? We can find drug-targetable mutations in tumors, we can direct patient care in rare (and some common-ish) diseases from germline mutations. Now, with liquid biopsy we can actually track tumor progress through treatment and identify resistance mutations that pop up. Before long, we'll be able to do cancer screenings from blood! None of that would be possible without cheap, scalable NGS sequencing tech.
Tumor sequencing is a HUGE boon to cancer treatment and research. Patients lives have been drastically improved by targeted treatment.
Targeted therapies were not developed as a result of genomics. Erlotinib was approved in 2005. We've had Sanger sequencing assays for EGFR, BRAF etc for years, and they're still standard of care. We can identify resistance mutations, but we often don't know what to do with this information. Your patient becomes resistant, you identify the causative variants, you watch your patient progress and die.
Sure, genomics gives us a comprehensive way to sequence, but my point is that it did not yield the sudden, vast sweep of insights that we expected it would.
IMO, it's a bit like geneticists walked into the human genome expecting a simple SQL database, and discovered the thing was littered with pl/sql functions, and that some of the functions actually double as data, when shifted over a byte or two. It's no surprise then that it's taking longer to understand the genome, and nor surprising why differential genomics is so useful.
Whether this was the best use of scientific resources, I donno.
While 23andme's terms dictate that you should use your own identity, in practice you could circumvent that by asking your friend to provide the info and credit card payment. Preferably same sex.
This is nitpicky, but 23andme doesn't do whole genome sequencing. They use a genotyping SNP chip that interrogates about 500k individual bases (and imputes other positions) rather than sequencing the 2x3B base pairs in the whole genome.
I would recommend talking to a local genetic counselor. They may be able to find a genomics core that will fit your needs, and they will be able to help you interpret that data in a medically relevant way.
I was under the impression that modern genomics had moved away from the "gender variants are what matters" mindset. Hasn't greater understanding of the proteom and the influence of the environment massively complicated this already daunting task?
I hope it brings with it some advances in diagnostics. I have some kind of autoimmune condition (I think) that has resisted a proper diagnosis for over 2 decades. Whatever it is, I lost my livelihood to it, and now I stand to lose my savings as well. Since I have no diagnosis I can't even apply for social security benefits, much less hope to return to work (I used to write code pretty successfully).
I assume you have checked for allergies/intolerances (e.g. just vegetables for two weeks and changing the physical environment around)?
I had effects for most of a week from my intolerances (nuts, peanuts and others), so I never really understood that I had them -- I got some allergen in me more often by pure chance.
The biggest effect for me was relatively mild compared to you -- sugar/caffeine overuse to stay awake and depression. These days, my cheerful disposition would have really irritated the old me. :-)
If you're interested in genomics, you should take a look at the pharmacogenomics knowledgebase from Stanford: http://www.pharmgkb.org/
It is a comprehensive database of dosing guidelines, drug labels, and annotations for thousands of combinations of genes, variants, drugs, and diseases.
(We're actually helping them with a redesign, and if you'd like to be a beta tester, contact me.)
I would highly recommend the book 'The Emperor of All Maladies'. When you read a detailed account of the fight against cancer two things become clear. First, we had no idea what we were doing for a very long time. Second, the era of actually fighting some of these more complicated diseases is upon us. We have been trying to stop cancer for nearly 100 years but only in the last 15-20 have we actually been making any progress. It's a well written book and makes me quite hopeful for the future.
There's very little meat to support the subhead's assertion.
In Robert Gordon's The Rise and Fall of American Growth, a great deal of attention is focused profitably on the changes and progress in medicine and health outcomes from 1870 to present. Most notably, Gordon divides the period at 1950, noting that life expectencies improves 2x more before 1950 than afterward (and from my own explorations, far more prior to 1920 than after).
There's been exceedingly little progress in medicine since 1970, at which point major cancer, heart disease, and virtually all infectuous disease treatments existed. We've seen many more treatments and more so imaging and diagnostic capabilities since ... but virtually no changes in outcomes.
Healthcare is a land of very, very, very rapidly diminishing returns, and for which shoring up treatment and preventive care for the most under-served pays hugely greater dividends than highly invasive or heroic treatments at the top end. Much of the progress in health outcomes since 1970 appears to be in minority populations -- that is, the under-served.
Gordon's book was published this year, its information is quite current. I see little reason to suspect massive improvements in actual outcomes -- impacts rather than change -- in the nine months or so since it was put to bed.
This is broadly false; outcomes have in fact improved greatly in very many fields. Off the top of my head: Far more people surive heart disease and heart attacks; AIDS didn't even exist in 1970.
I'm only half agreeing with you. Yes, improvements in medicine before 1920 were dramatic. Antibiotics and vaccines work wonders, making once-deadly diseases benign (e.g. pneumonia) or eradicated (e.g. smallpox).
But more recent improvements in healtcare aren't just trivial stuff either. The five-year survival rates for childhood leukemia, for instance, have been on a steady increase from 40% to 80% since 1970. Recent monoclonal antibody therapies against inflammation/autoimmune diseases have provided actually revolutionary changes in the health situation of many.
Going from chronic pain and having a hard time functioning at school/work, to living a relatively normal life, is a huge thing. But it's not captured by life expectancies.
83 comments
[ 4.1 ms ] story [ 147 ms ] threadUpdate: Please ignore my comment. The original post was changed and my upvotes have turned to downvotes. Now I can't delete this comment.
The costs of assaying a genome has dropped from $100 million to $1 thousand.
Identifying correlations with clinical metadata, understanding the post translational interactions and understanding the full roles of proteins and other molecules remains an intimidating and difficult job.
http://www.genomicsengland.co.uk/the-100000-genomes-project/
Edit: I realize it's a small part, but it's the first step, and it's already relatively inexpensive, which is a good sign.
I'm not saying that it's not worth it in some cases, but the sequencing is only a small part of the overall costs.
Interpretation, on the other hand, is quite different. Currently, interpretation still requires a human to look at everything. And regardless of how good annotations and algorithms get, I have yet to hear of anyplace where the final annotations aren't manually curated (maybe automated, but still requires a human in the loop). This aspect of Precision Medicine doesn't scale quite as well as sequencing or computation...
I'm not trying to belittle recent advances in genomics, on the contrary I believe that cheap sequencing and technologies like CRISPR are going to change the world. However, its the insinuation that the last 20+ years of hard work were a fraud that upsets me. We wouldn't be were we are now without the last 20 years of genomics research. /rant
Actually, cells have a lot of state. See [1] for a basic, slightly old, CS-biased explanation.
That's how a single genome yields an incredibly varied repertoire of cell types. We need to work a lot on understanding those epigenomes to then get a chance to pinpoint genetic causes of disease.
[1] http://ds9a.nl/amazing-dna/
As someone with a CS background who has worked almost 5 years in bioinformatics, one thing to realise is if you think something obvious is not done, either you're missing the right search term on pubmed or there's some horrible lab issue that makes it really hard (and we need a new assay/tech)
A few decades ago ago we read our first line of source code from a computer from the year 2,000,000,000 - some pretty advanced technology that was just handed to us. Two decades ago we did our very first read-through of an entire human. The reading itself was not particularly productive. We already knew some of the most important lines of code, from the decades before, and we annotated a lot more sections that are critical for basic function. We've even traced a few bugs back to the source code - especially those bugs where 1) the source was unusually easy to read [1], or 2) the bug was really bad [2]. As of this year we've developed compilers to enable us to write and run small, multi-line programs of our own composition [3]. We're doing really well in trying to read, comprehend, manipulate, and act on some of the most sophisticated technology ever presented to us. More sophisticated by orders of magnitude than the silicon in Nvidia's new graphics card (16nm 2d dry feature size [4], vs 0.1nm, 3D wet feature size [5]) . Reverse engineering it takes a bit of time, and yet we've already put that knowledge to significant use.
Further, once we do get our working knowledge up and running there will be a significant outpouring of function from having reverse-engineered such technology. But of course there is a lag time in the millions of lab-hours being poured into understanding that technology before we can actually use it. Not dissimilar to the rapid advances in science-fiction when a race reverse-engineers an advanced alien technology. How long did it take us to build The Machine in Sagan's 'Contact'? We will not just cure cancer, but we will be capable of all sorts of new feats. Purchasing the textbook doesn't make one a master - it's the study, the hard work, and the practice.
[1] https://en.wikipedia.org/wiki/Human_papillomavirus
[2] https://en.wikipedia.org/wiki/BRCA_mutation
[3] https://news.ycombinator.com/item?id=11417689
[4] http://wccftech.com/rumor-nvidia-pascal-gtx-1080-gddr5x-gtx-...
[5] https://en.wikipedia.org/wiki/Resolution_%28electron_density...
freudian slip? ;-)
Do I need to dig out this story or do you mean Carl Sagan's 'Contact'?
For a scientist working incrementally, accumulating data, sieving it for information, fighting for grants, doing administration while making sure family gets three squares, articles like this may cause a wince. But they probably influence support for more funding. Pro scientists I know do not have time to read these articles and will never see them unless it's from someone else, but budding scientists, career-changers, influencers, all get misinformed.
In short, news on science causes long-term damage for short-term gains.
BLAST (Stephen Altschul, Warren Gish, Webb Miller, Eugene Myers, and David J. Lipman)
BWA (Heng Li, Richard Durbin)
Samtools (Heng Li)
GATK (from the GATK Team; DePristo, M., Banks, E., Poplin, R., Garimella, K., Maguire, J., Hartl., C., Philippakis, A., del Angel, G., Rivas, M.A, Hanna, M., McKenna, A., Fennell, T. Kernytsky, A., Sivachenko, A, Cibulskis, K., Gabriel, S., Altshuler, D. and Daly, M. A)
I'm missing quite a few from my more niche field; but feel free to add more.
http://www.1000genomes.org/
The VCF files are reasonably formatted, the raw sequencing data is in the FASTQs, which are huge and hard to deal with. Go nuts!
Here is the consent template: http://www.1000genomes.org/sites/1000genomes.org/files/docs/...
The post I responded to sounded like the genomic data would be uploaded without consent to a giant public database.
Even with consent the greater concern is that employers, medical providers, national health services, and health insurance companies could, by sequencing part of your genome, match it to a public database identifying you and your entire genome. This could then be used to raise prices, deny health care, or deny employment.
De-identification is not enough if the material is indicative enough that efforts to re-identify it can succeed.
Kudos to those advancing science by making their genomes public, but it's a risk I would not take.
With new technology comes new responsibility - we shouldn't shy away from the tech because we don't want the burden of enforcing regulation.
For me, the big story in genomics over the past few years is one of massive failure. We spent a lot of time and effort sequencing tumors in the hopes that the genetics would tell us something interesting and lead to cures. It did not. We now have a relatively complete cabinet of the major variants that drive tumors. Most of these variants we already knew about before we did these genomics studies (through older sequencing methods from decades prior), and most of what we learned tells us nothing new.
Genetics is at this point mostly garbage information. Why? Because we don't know what it means. We still don't understand gene expression, we don't understand signal transduction, and we can't understand the effects of mutations without painstaking characterization. All of these things mean our sudden wealth of knowledge of genetic variation tells us fuck-all about biology.
We learned quickly that 'cancer' is to the twentieth century 'the fevers'. We learned there are many reasons and causes for the catch-all, 'cancer'. Some we now know how to cure effectively. Some we now now how to prevent, effectively. Some we now know what to do to cure, but don't have good tools yet. And many we still do not understand. That is a significant improvement in a very short time span.
But there's a LOT more to learn than 'how to cure cancer'. We very clearly understand gene expression and signal transaction to a first order, and have bits of the second order down too. There might be more - but that doesn't make that first order inneffective or wrong. There's lots you can do with a first order understanding - see what Newtonian physics did for the world - it was correct only to the first order. The idea that we know 'fuck-all' about biology because of dna sequencing is just ridiculous. The ability to read and write dna is fueling one of the fastest growing segments of human progress today.
The proof in the pudding is this: you can get your exome sequenced today and report hundreds of tumor variants, but there is a bare handful of variants that will result in a change in your treatment, and most of those lesions were well-studied before genomics took off.
>We very clearly understand gene expression and signal transaction to a first order, and have bits of the second order down too. There might be more - but that doesn't make that first order inneffective or wrong.
Yes, in fact, our first order understanding IS ineffective. This is what I study every day, so perhaps I'm too close to this, but we literally do not understand the effect of most (99.9%) of genetic variants on gene expression. Sure, there's lots you can do with a first order understanding, but materially, what happens when you sequence a tumor (or a germline, for that matter), is that researchers get a list of mutations, stare at it mystified for a while, and then shrug and move on, because there's really nothing you can do to understand what these things mean.
Yes, this stuff is the way of the future, and it's important to do it for our greater knowledge and understanding in the future. But we're far, far away from this point right now.
I think there's probably a fifteen-year lag in this stuff being useful. The insights we're having now are because of the sequencing of the human genome; the work we're doing now will pay off in another decade and a half when we've figured out how to deal with gene expression.
>do the full genome sequences yet actually read every base in your genome even?
Read, sequence: yes. Align: no.
They are not "blueprints" much less some sort of engineering document.
More realistically, they are a simple list of materials (in this case a listing of proteins).
They only describe the materials that make up a structure, not what that structure is nor how it functions.
Its the equivalent of someone saying: 14,234 tons of steel, 23,000 tons of concrete, 8000 tons of glass...etc. Now, what does it make? Why does it make it?
I think of genes as the initial configuration in Conway's game of life. Small initial (genetic) variations can cause large differences after some generations. Some don't matter at all and just die out. But no matter what is the end result, it's not a blue print or a description, it just _is_, and the result flows from it using some basic rules.
https://en.wikipedia.org/wiki/Epigenetics#Transgenerational
Epigenetics as a concept can also be abused into a sort of neo-Lamarckism.
But good luck growing a healthy organism without its epigenetic landscape properly initialized
This is beside the point, as parent already explained what he meant in his first 2 sentences: "Physics. More specifically, 'conditions'".
So, yes, the laws of physics might be universal, but gravity is X here and N on the moon, pressure is Y here and M under the sea, temperature, etc...
So according to one definition, it's even worse than your 14,234 tons of steel example: it's closer to "steel is a mixture of iron and carbon and trace amounts of other dopants, glass is silicon dioxide, concrete is ..." Without even listing the amounts. But by another definition, the genes are, if not a blueprint, at the very least a recipe and ingredient list. "Butter, flour, water, sugar, apples, cinnamon, cloves. Cut 125g butter into flour; if too coarse, continue cutting. Add water to ice. Add water from ice into flour..."
The information is clearly there—we have a long, unbroken history of offspring being very similar to their parents, despite having started with only a single, undifferentiated cell. But with the new tools available to us, we're starting to have a hope of understanding it, even if we're using more brute force than we'd like.
> Its the equivalent of someone saying: 14,234 tons of steel..
I feel like that's taking things a bit too far in the other direction. With our current understanding, it's like saying:
14,234 tons of steel that's capable of self-assembling (e.g., collagen), 23,000 tons of concrete that automatically folds into a specific structure (~everything other than disordered proteins).
That's not to say that we fully understand how proteins fold/function... but it's a bit more than a static block of steel.
Tumor sequencing is a HUGE boon to cancer treatment and research. Patients lives have been drastically improved by targeted treatment.
Sure, genomics gives us a comprehensive way to sequence, but my point is that it did not yield the sudden, vast sweep of insights that we expected it would.
Whether this was the best use of scientific resources, I donno.
I had effects for most of a week from my intolerances (nuts, peanuts and others), so I never really understood that I had them -- I got some allergen in me more often by pure chance.
The biggest effect for me was relatively mild compared to you -- sugar/caffeine overuse to stay awake and depression. These days, my cheerful disposition would have really irritated the old me. :-)
It is a comprehensive database of dosing guidelines, drug labels, and annotations for thousands of combinations of genes, variants, drugs, and diseases.
(We're actually helping them with a redesign, and if you'd like to be a beta tester, contact me.)
In Robert Gordon's The Rise and Fall of American Growth, a great deal of attention is focused profitably on the changes and progress in medicine and health outcomes from 1870 to present. Most notably, Gordon divides the period at 1950, noting that life expectencies improves 2x more before 1950 than afterward (and from my own explorations, far more prior to 1920 than after).
There's been exceedingly little progress in medicine since 1970, at which point major cancer, heart disease, and virtually all infectuous disease treatments existed. We've seen many more treatments and more so imaging and diagnostic capabilities since ... but virtually no changes in outcomes.
Healthcare is a land of very, very, very rapidly diminishing returns, and for which shoring up treatment and preventive care for the most under-served pays hugely greater dividends than highly invasive or heroic treatments at the top end. Much of the progress in health outcomes since 1970 appears to be in minority populations -- that is, the under-served.
Gordon's book was published this year, its information is quite current. I see little reason to suspect massive improvements in actual outcomes -- impacts rather than change -- in the nine months or so since it was put to bed.
But more recent improvements in healtcare aren't just trivial stuff either. The five-year survival rates for childhood leukemia, for instance, have been on a steady increase from 40% to 80% since 1970. Recent monoclonal antibody therapies against inflammation/autoimmune diseases have provided actually revolutionary changes in the health situation of many.
Going from chronic pain and having a hard time functioning at school/work, to living a relatively normal life, is a huge thing. But it's not captured by life expectancies.