How Big Data Can Help Fight Cancer
cancer.nautil.us
cancer.nautil.us
As nice as Foundation's data is, there are few/no phenotypes to go along with it, representing a bottleneck to extracting any useful information out of it. Keep in mind that genomics is only one half of genetics (the other half being the phenotype) and is meaningless without knowing more about the patient.
On the phenotype side, EMR/EHRs are practically useless as scientific tools (doesn't prevent academics from publishing how they extracted meaningful data from it.. which has little correlation with if something works or is reproducible; unfortunately) and I don't foresee this being fixed in the current healthcare ecosystem in the US. The UK & other European governments with good data + national healthcare systems represent a better chance.
If you don't believe me, I challenge you to point me to a single study containing a comparison of 1000 metastatic sites vs primary tumors in one cancer type, with WGS. Given that metastasis causes > 90% of deaths (ref: Weinberg cancer textbook), you'd think we'd have done this study by now.
A dx company has no hope of reimbursement for doing cancer whole genomes, and does not receive patient data in sufficient detail (how did they fare after treatment? what drugs were they given?) to undertake a study.
As it is, I fear any and every bit of data I provide the system may well be used against me at a future point.
Right now, I'm going through some extensive testing, and I've decided to provide further historical data in my possession for the sake of a better analysis and diagnosis. However, that same data -- or rather, one datum of the data it is comprised of -- a few decades ago, was used as the basis to deny my application to purchase individual health care insurance.
With the ongoing attacks on the Affordable Care Act, I have the distinct feeling of traveling back in time.
If we are going to have cooperative buy-in on big data, we are going to need to ensure that the resulting benefits are shared across the population and are not used to discriminate against subjects having data "on the left side of the bell curve".
"Nevertheless, the rate of chromosome missegregation in untreated RPE-1 and HCT116 cells is 0.025% per chromosome and increases to 0.6 – 0.8% per chromosome upon the induction of merotely through mitotic recovery from either monastrol or nocodazole treatment ( Fig. 3 C ). These basal and induced rates of chromosome missegregation are similar to those previously measured in primary human fibroblasts ( Cimini et al., 1999 ). Assuming all chromosomes behave equivalently, RPE-1 and HCT116 cells missegregate a chromosome every 100 cell divisions unless merotely is experimentally elevated, whereupon they missegregate a chromosome every third cell division. Chromosome missegregation rates in three aneuploid tumor cell lines with CIN range from 0.3 to 1.0% per chromosome (Fig. 3 C ). Depending on the modal chromosome number in each cell line, these cells missegregate a chromosome every cell division (Caco2), every other cell division (MCF-7), or every fifth cell division (HT29)." https://www.ncbi.nlm.nih.gov/pubmed/18283116
Many people claim that aneuploidy is found in nearly all cancer cells:
https://www.ncbi.nlm.nih.gov/pubmed/17046232
I think the problem is big data, in a non statistican sense, where you cannot get large amount of observation from patient. Either because of legal loop holes and/or the cost of getting enough patients for experiment and trials. This is for trials.
Even with phase III clinical trial it's less than 200 obs. This is not big data in the non statistician world. In our world big data mean tons of predictors. Medical data is usually high dimensional, less obs but tons of predictors, more columns than rows.
Also hospitals are wary of giving out data, either because of legal issues or because they know it's valuable so they don't want to share it. These two issue is compound it on the fact that the infrastructure is not there to share the data in one spot, it's fragmented across many other databases with different schema and what not.
But my peers and I have thesis involving cancer using genetic data. It's very promising, one of the recent thesis is about base on genetic data if the patient should take the surgery route or the chemo route and the model had a 80% accuracy rate and nice sensitivity rate (forgot what it was). The prediction is survival rate.
I also saw other comment about using genetic data against them. I think this is FUD because we have a law in place, GINA.
Additionally, one of the greatest revolutions in cancer therapy - immunotherapy - does not have a great genomic-based predictive biomarker (Foundation Medicine has created a surrogate test using mutation burden, but this is not really good enough, and not validated prospectively).
So DNA sequencing is only part of the story. Some tumours in particular seem driven by epigenetic changes, or structural variants that Foundation cannot detect.
Sequencing is in a place where it has reached technical maturity, and has collided with big data hype. But in reality, the benefit to patients will be incremental. There is at present zero good quality evidence that panel based genomic screening like Foundation Medicine provides improves patient outcomes as a general strategy.
[1] https://www.genome.gov/sequencingcosts/
[2] http://www.nature.com/nrg/journal/v17/n2/abs/nrg.2015.17.htm...
That's not entirely accurate. The majority of variants occurs in the non-coding genome, but that's also because the vast majority of the genome is non-coding. There certainly are non-coding variants that are oncogenic (e.g. hTERT promoter), but for the most part the functional significance of a given non-coding variant is unknown.
I'm not suggesting greater sequencing isn't important, but there are a lot of considerations that goes into what part of a cancer genome get sequenced (WGS, exome, gene panels, specific variants). We simply don't know what a lot of the variants do, but for those that suggest a particular therapy (BRAF V600E), it can be quite effective.
1) Exome sequencing is below $1000, but analyzing that data adds a non-negligible cost. Still, even 2 or 3 grand is way cheaper than wasting time on treatments that won't work. Whole-genome sequencing is even better (for a little more cost) because of the extra types of information it adds about structural variants and copy number changes.
2) We're reasonably sure that most cancer causing variants are in the coding space, but there are undoubtedly some in non-coding regions (and classes of large structural events like duplications or deletions that affect both).
It's the best time in the history of the world to have cancer, and it is only getting better. Survival curves are slowly bending, and new classes of treatments like immunotherapies are helping to bend them even more.
The bottom line is, If you get cancer, fight like hell to get your tumor sequenced. Most insurers cover at least some kind of genomic test for cancer these days.
What are you basing this on? Have you seen a study where they report "modification" in a living organism using this tech? I highly doubt it, due to toxicity. Also see my post above regarding the presence of aneuploidy and chromosomal instability in cancer cells. How is crispr/cas9 going to fix that?
Logical extension of how the strategies for technologies like Car-T therapies are being developed and delivered. By first targeting an ex-vivo tissue (like T-cells that can be harvested and then re-implanted), you first develop the tools to use genome editing techniques effectively, without significant off-target effects. It's essentially practice for much more useful therapies that can be done in-vivo. Further, the design of the biological 'sensors' in the T-cells are precisely the kinds of sensors you'd need to deliver a payload to and only to a mutation-containing cell. So you would neither need to target an entire organism (rather a tissue, or even a particular cell type), and off target effects from genomic insertion would be curtailed by advances in the technologies surrounding Cas9 itself. For many cancers you wouldn't need to repair the mutation in the entire organism as the cancer is tissue- or cell-specific.
You are correct that you would, however, likely need to repair the prior to significant metastasis or loss of stability of the chromosome. At that point fixing a point mutation is not going to help much. All the more reason screening and baselines should start being established now - so that we can even detect a mutation prior to it becoming 'cancerous'.
If you know you have a stop codon mutation in your Her2 gene when you're born, it will not be too far off when that stop codon, in particularly sensitive tissues, could be reverted with reasonable safety prior to the development of an (inevitable) cancer. And ultimately that actually cures the the cancer in a way a small molecule can never, from first principles, hope to do.
Here is actually one where they injected into Drosophila embryos. It looks like to get 10% success rate, they had to kill 50% of the "subjects" (table 1): https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3714591/
Here is another where you see the more cas9 you inject, the more animals (C. elegans) die (table S1): https://www.ncbi.nlm.nih.gov/pubmed/23979586
Further, the body is pretty resilient. You don't need viability to be 100% before a therapy could be successful. Chemotherapy kills a whole lot of cells and people survive it.
Entirely new (smaller, thus more engineerable) Cas systems (1 month ago): http://www.nature.com/cr/journal/v27/n3/full/cr201721a.html
Variously engineered Cas9 proteins to reduce toxicity the over wild-type protein (two months ago): https://www.ncbi.nlm.nih.gov/pubmed/28179977
Cas9-Inhibitors to help regulate activity (two months ago) http://www.cell.com/cell/fulltext/S0092-8674(16)31683-X
And this doesn't include any ongoing research using other homing nuclease technologies like TALENs and Zn-Fingers that provide an entirely different set of tradeoffs but aren't as sexy, in part because they're harder to use.
Also, I think if they were serious about addressing this toxicity issue they would be include simple charts of the number of cells over time and number of "modified" cells over time. Then they (and the reader) can easily compare the toxicity and efficacy of different techniques.
That is by and large the most significant factor that we've seen which slows adoption of more widespread whole-exome or whole-genome sequencing. It takes less than a day and costs less than $1000 to sequence your exome (and even your whole genome), but the backlog for analysis of the sequencing results in labs can be 9 months or more.
There are patients that could be treated from the analysis of their genome that aren't even considered if their prognosis is shorter than the amount of time it will take to get clinically actionable results from the lab, which is the really sad part. This problem is actually what led us to create our company, Genomenon, to help make the analysis much faster and alleviate the bioinformatics bottleneck at the sequencing labs.
Layman qs. what is stopping the labs to quickly analyze the genome? Computational power or few labs doing this kind of work?
Imagine you want to reconstruct the data on two RAIDs that are mostly, but importantly not exactly, mirrors of each other. Each RAID has 23 drives. Each drive has ~1Gb or so of data. And much of the data is not only mirrored between the two RAIDs, but is also mirrored between the 23 drives - and many of that mirrored data is 'off by 1' in very important ways (both 'must', and 'must not' scenarios). Further some of the data contains very long sections of highly repetitive data. And some of the data is mechanically biased to be harder to read than others.
You must now reconstruct those two RAIDs with single-bit accuracy - as a single bit-flip in certain sections determines whether or not you get cancer. The data you are given to do the reconstruction is a 200Gb single column CSV file with each row being 12 bytes of data.
Go.
In addition to our company, Genomenon (which seeks to speed up the interpretation time required to analyze the data _after_ it's been computationally aligned and annotated), I'm also friends with another startup down the street, called Parabricks, which seeks to speed up this alignment process (aka secondary analysis) even further.
If you are further interested in parallelization schemes for genomic pipelines, please have a look at our paper on the strengths and limitations of big data technology for genomic analysis (published last week) - https://people.cs.umass.edu/~aroy/sigmod17-roy.pdf
Something worth thinking about: Processing on HIPAA-compliant resources is expensive, and as with any medical procedure, that financing/insurance is complicated to bill for (or at least, takes a long time... which makes it more expensive to run, etc).
The part that takes a long time (i.e. the "bioinformatics bottleneck" I referred to), is that once the sequencing data is stitched together, you end up with a ton of variants, and you don't know which (if any) are clinically significant.
Imagine that each nucleotide in your genome is a marble, and that the entirety of your sequenced genome is a 1-story building filled with marbles (that's how many nucleotides are in your genome), and that each one is supposed to be a specific color out of 4 possible colors. Now imagine that 10,000 (or more) of those marbles are the wrong color.
Primary analysis (putting your sample into a machine and essentially getting back a list of what color all your marbles are), as well as secondary analysis (i.e. the process of sorting your marbles into the correct order so that you can actually tell _which_ marbles specifically are the wrong color) together are what cost less than $1000 and take less than a day.
The real problem is that you may find that 10,000 (or more) of your marbles are the "wrong" color, but 9,900 of them make absolutely no difference in a clinical sense. To be clear, a marble that's the wrong color is a "mutation", or "variant". Maybe this mutation makes my eyes slightly bluer, or my finger nails a little harder.
In other words, the part that takes a long time is actually going through each mutation and figuring out which one (out of the 10,000 variants) is clinically significant or relevant to the disease/symptoms you seem to have, and then figuring out if there's a known treatment for that particular root cause. Currently, this is done by employing MD PhDs to look through each patient's sequenced data, and then cross-referencing that with the millions of published studies to see what has ever been seen before, or is known to be associated with some disease.
And it can take a human hours to a day to do this per patient. So, the number of patients times the amount of time it takes per patient, divided by the number of MD PhDs a lab can hire to do this, is what leads to the backlog.
So, actually yes, it is computational power... but it's human computational power.
However, it's far from perfect and can always be improved. That's one of the things we're trying to help with, is to do further pre-processing on the knowledge-base of all genomic information contained in all published literature, to try to drive the current hours-to-a-day timeframe down to minutes-to-an-hour, or just minutes.
One issue though, is that it's a constant tug-of-war optimizing between false-positives (giving you too many variants to manually review) and false-negatives (removing variants due to some data threshold which may have been clinically significant).
Is this something that has to go through your oncologist or can you drive the process independently?