chrisamiller

I don't mean this to sound snarky, but if a blog doesn't have a good ratio of signal to noise, you just unsubscribe from the feed.

The solution is to be okay with missing some things instead of trying to drink from the firehose.

Our middle school has banned phones during school hours, and it's been great. Even though most kids own phones and have them in their backpacks, they aren't seen. There are less problems during classes and kids have to talk to each other during lunch instead of going head down in their phones. Parental pushback has been really minimal - most parents love it, and it's good for the kids.

Making no phones a blanket policy in all schools should be an absolute no-brainer.

At any given time, I'm working on like 10 different projects. tmux lets me set up sessions for each of these projects, so that when I leave one and come back in a week, all the context is there (multiple windows, bash history, working directories I was in, etc)

Since I work with big biological data, most of my work takes place on our university cluster, which means my laptop is just a dumb terminal and all of the action takes place on the server. IME, tmux is especially powerful with coupled with mosh, which gives a persistent SSH connection. That means I can be in the middle of a project, close my laptop lid, go home, then later that evening, open my lid and everything is reconnected and just right there. Same if I reboot my laptop - one command to reconnect my terminal with mosh, and I'm back in the middle of my complicated multi-window project.

Are you claiming that declining fertility is due to plastic, rather than to increasing wealth, education of women, and access to contraception? If so, then I'm going to have to ask you for a) a source and b) a plausible mechanism.

Absolutely. We're not there yet, but it's not inconceivable.

The short answer is, we sequence their genomes, identify mutations that change a protein sequence and are highly expressed, then run those all through suites of algorithms that predict how well they'll be presented to the immune system. (mostly neural networks trained on far-too-sparse experimental data). This prediction is the hard part right now - we still don't understand enough about how the immune system identifies and interacts with these altered peptides to do really accurate predictions of which ones will be most effective. Throw in that these tumors are actively suppressing the immune system in various ways, and it's complicated! There is lots of research going on, though, and lots of promising early results, like this one

Look, I get the impulse, and might have agreed with you 10 years ago. At the end of the day, though, we have to work with non-computationally savvy people who (reasonably) want to look at their data sometimes. Not every lab tech or PI can or should learn Python/R/your favorite scripting language, and frameworks like Galaxy take time to set up and maintain, etc etc. Our job is to meet the users where they are and push the biology forward. Any time that I can push them onto a better path, I'll do that, but sometimes the right move is to tell them: "Yeah, go ahead and play with the data in excel, then tell me what you find", and I'll code something proper up afterwards to verify it, get solid stats and a make a pretty visualization.

It's too late, the genetics community already caved :-(

[Scientists rename human genes to stop Microsoft Excel from misreading them as dates](https://www.theverge.com/2020/8/6/21355674/human-genes-renam...)

These are ballpark numbers to be sure, and yeah, my off the cuff comment didn't get that exactly right. I also simplified things quite a bit to try to get the broader point across - Thanks for following up!

This is a great example of where cancer treatment is headed and why it's so hard - namely that cancer isn't one disease, it's many thousands of diseases.

This is a drug that targets lung cancer (~12% of cancers) and only one type of lung cancer (non-small cell lung cancer, ~80% of cases). It targets a particular mutated gene that occurs in about 30% of that subtype. And then, about 50% of those patients respond.

So do that math, and you end up seeing that treating one of the most common mutations in one of the most common cancers with what is considered very high efficacy still only helps with about 1.4% of all cancers. This is actually an enormous number for this kind of treatment, and there is a long tail of rare cancers that are going to be much harder to find targeted therapies for.

That all said, this currently appears to be an enormous success story, and the kind of treatment options that have been enabled by genomic sequencing of cancers, followed by many years of drug development and clinical trials. It's fantasically exciting to see us continue to chip away at the problem but by bit and grant people longer lives as a result!

I do cancer research, including using mouse models, so believe me, I know the difficulty :-)

Getting promising drugs into humans is -hard-, and this research is interesting, but this press release is still kind of shitty.

I hate to be that guy, because it seems like there is some interesting science behind this press release hype. There are, however, many many miles between "effects in a mouse model" and "human therapeutic", let alone "cure". I wish them the best of luck as they consider safety and efficacy in clinical trials.

You're going to want to start with the PCAWG consortium, who looked very closely at viral involvement across a host of cancers. Their numbers:

Cohort: 2,658 cancers across 38 tumor types Findings: Overall, 23 virus genera were detected across 356 patients with cancer (13%)

You can read all of the details here: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8076016/

I'm not saying that there aren't other mechanisms (sustained inflammation, etc) that might contribute to the aetiology of some other cancers, even without clear viral integrations, but we can state pretty strongly that many cancers are not directly caused by cancer.

It's true that the whole problem of cancer therapy is "how do I target these cells over here, while avoiding the rest of your healthy cells, when they look mostly the same" And you're right that this is a really exciting and promising approach!

A partial list of the "catches" at present are

1) tumors are really good at down-regulating the immune system. If they weren't, they would have been cleared already

2) we're still fairly bad at knowing which of the many mutations in a tumor are going to be good neoantigens - that is, which ones the immune system will be good at recognizing and ramping up against

3) It's expensive to design personalized therapies for every patient, but may be necessary (see some of my other comments below)

4) Tumors aren't single cells, they're collections of them that can have different characteristics. Just like with other therapies, sometimes immunotherapies kill 99% of the tumor, but leave that 1% to come roaring back. (same idea as antibiotic resistance)

So yeah, lots of hurdles to overcome, but overall, I'm genuinely excited about the rapid progress we're making in the area of immunotherapies. It's impossible to overstate the effect that sustained research funding has had on the rate of advances in this field.

There is a lot of great work going on thinking about this very problem. One common theory goes like this:

There are lots of tumor "driver mutations" that are causative and occur in lots and lots of patients. These would make great targets for a vaccine like this, in theory.

But - Tumors have to evade the immune system in order to gain a foothold. (We have all had many cancerous cells that were cleared away without anyone ever knowing!). If a mutation is particularly good at activating the immune system, it's unlikely to survive long enough to cause cancer. So one of the features that a mutation has to have in order to become a driver is that it's bad at being recognized by the immune system! That means the most common mutations that we'd like to target with cheap, mass-produced vaccines may not work very well.

There's some evidence on both sides at this point, with lots more data being collected, but it's certainly something that us folks researching the problem are keenly following!

I'm as critical of pharma pricing as the next guy, but this really is a case where you're getting a custom vaccine design just for you, and the profit margins probably aren't very high. I certainly do think it'll come down eventually, but there are lots of factors slowing that down.

This isn't a preventative vaccine like that. It's something that has to be developed after you have cancer to target your tumor's specific mutational profile. So mass production (just churning out the same thing in bulk) isn't really an option! Costs will probably come down eventually, but it's going to take research, technology development, and hard work to make that happen.

That's not far off from the rates for producing custom order medical-grade peptides and/or mRNA vaccines, coupled with the expertise needed to design, administer, and monitor them. It's certainly not a case where they produce it for 100 dollars and sell it for 100k. The market for this stuff is beginning to develop, and I hope that we'll see new innovative companies begin to produce them more cheaply.

There are lots of folks working on problems surrounding automation and reproducibility to enable science in this space. One such effort that I contribute to is [pVACtools](https://pvactools.readthedocs.io/en/latest/), (there are others as well). We do have a small team of developers who work in concert with scientists and clinicians to build these tools.

You're right that there are a variety of folks working on these things, both in academia and industry, but it's more complicated than "PhDs do this, others do that". It comes down to knowledge, ability, and experience, and the best research groups grab onto that wherever they can and slot people into the roles that are right for them.

Part of my research involves designing these cancer vaccines. There are different delivery mechanisms, but the need to synthesize medical-grade peptides or mRNA vaccines is one of the major drivers of cost. The sequencing costs are pretty minimal at this point, but analysis of that data by a team of experts is part of that cost as well. This is still largely bespoke stuff that can't quite be standardized because we're still figuring out the best approaches, but there are quite a lot of people working on it!

The amount of energy needed to get out of earth's gravity well is incredibly high. It would be silly to try to boost a massive object like the ISS out.

That's a pretty expansive statement. Are you doing error-corrected sequencing to look for rare mosaic variants? Do you care about transcripts expressed at very low levels? Are you trying to tease apart subclonal variation in cancer? All of these could be good reasons to sequence deeply.

If anything, I see the opposite problem much more frequently: People who cheap out on the sequencing depth, neutering their statistical power!

In most cases* the raw data can be recreated from the aligned BAM/CRAM file, so it's often sensible to toss the raw FASTQs after alignment. But yeah, I go back to 10+ year old genome sequences more than some might expect.

*In the absence of read trimming or discarding unmapped reads

Money is absolutely an object. People are starting to do trials of early detection from circulating tumor DNA, and finding that the hit rate is alarmingly high. A decent portion of the population has some kind of malignancy, but most of these will either be cleared by the immune system, or be completely benign. Another problem is, if we detect this tumor DNA, it doesn't tell us where in the body the tumor might be. So then you're talking massive workups including while body CT scans and such for a problem that isn't going to be a problem for 99% of people. I'm not joking when I say that if rolled out as is, this has the potential to bankrupt the entire healthcare system with unnecessary procedures.

To be clear, I very much think that early detection of tumors will someday be an essential part of cancer treatment. We are not there yet.

I'm under no illusions that everyone is suddenly going to sacrifice quality of life to save the planet - humanity, writ large, is happy to burn coal until they choke to death on the fumes.

So how do we stop the slow-motion disaster that we're living in?

1) I completely agree that trying to minimize pain by teching our way out is important. The reason we have cheap wind and solar power is heavy government investment in research, which eventually trickles out and becomes commercially viable. The recent energy bill is great, but doesn't go nearly far enough. Still, if you take an honest look at the numbers, it's simply not going to be enough to avoid the worst effects.

2) We have to make it more profitable to be low-carbon than it is to burn things. I think that unavoidably leads to the conclusion that we need to tax carbon. It also has to be done in a smart way - not super-high all at once, and then return the proceeds with checks to the middle/lower income people who are hit hardest to solidify political support. The goal is to nudge people to change their behaviors, without doing it so hard that they vote you out of office.

Look, we were happy to keep dumping arsenic in streams until those externalized costs were built into the cost of doing business. The effects of carbon as a pollutant are here now - we can't keep pretending that they don't exist. What is the cost of Miami being underwater? What is the cost of wildfires and drought in the west? Yes, it will cost billions of dollars, but if it saves trillions, that's a smart investment.

There is already an unstoppable climate catastrophe happening. Many of the effects are already baked in for the next 50-100 years and it will disproportionately affect the poor.

What we can still control is how bad it is. (An increase 1.5 degrees C looks a hell of a lot different than 4 degrees C). That's still very much a fight worth fighting.

When people ask if we should do this or that, the answer should be "yes". These rules are fine - we should eat less meat, we should drive and fly less, etc. We should also do more systemic things, like investing heavily in battery tech and solar and wind and even fusion longshots. We should regulate the hell out of emissions, and use the proceeds from taxes and fines to help mitigate the effects on the poor. Getting to net zero carbon is going to be hard but it has to happen.

This story lacks a lot of context. This linked article provides some of it:

https://forbetterscience.com/2020/10/07/gregg-semenza-real-n...

There's a big range between "grad student loaded the wrong file when making sub-panel 3d" and "explicitly falsified images to support an hypothesis". This article certainly makes it seem like the latter.

Science is a human endeavor and there will always be people trying to game the system The good news is, an awful lot of cheaters get caught these days and it's getting riskier all the time.

But that was after an embargo period of 12 months, during which a journal could paywall it. This forces immediate availability, which is a good thing.

It will force publishers to either add real value or be swept away by new models of publishing that aren't simply rent-seeking. It will take time to change, but this is another hole in the dike. It'll probably be messy for a couple of years, but I welcome the opportunity to shake things up.