Li Tai Fang

Image shows PDF.

A blood-based test demonstrated acceptable accuracy for colorectal cancer detection in an average-risk colorectal cancer screening population.

#ASCO25

https://ja.ma/3HjVVoO

YouTube video by Anders Puck Nielsen Trump humiliates Zelensky to save his doomed peace process

My thoughts on the disastrous meeting between Zelensky and Trump. Trump is beginning to realize that his peace plan is failing, and he tried to bully Zelensky into submission. It didn't work.

Today, Trump & Vance cast a shit stain on the USA while doing the bidding of a "KGB war criminal" who has some kind of magical spell over the pair of them

⏵ Video

🇸🇰Slovak math olympiad medalist Simon Omaník declined to shake hands with pro-Russian President Peter Pellegrini during a ceremonial award presentation.

Omaník, wearing a Ukrainian flag ribbon on his suit, accepted his bronze medal from Education Minister Tomáš Drucker but subsequently refused

This is the real issue using sea water. They don't have the ability to draw it in large quantities, not because they don't want to because of corrosion or environmental concerns.

Corrosion takes a while to kick in while there is a $100B urban fire going on. The cost-benefit isn't even close.

Salt damage to the roads? Did you see fire damage to the roads?

I bet you don’t do that for every paper you’ve come across.

This. This THIS

We just need to give putin a..... what's that again... off ramp?

Great episode. I was very confused about this derogatory term “fishing expedition” when I started working with biologists. Was Galileo on a fishing expedition when he discovered the moons of Jupiter?

Make Apartheid Great Again

If you're a pharma coming up with a cure for cancer, it would be REALLY bad for every other pharma who just lost ALL of their revenues because they don't have a competitive product.

This is inevitable. Waymo vehicles are much safer than human drivers.

Targeting a senior military officer engaged in active hostilities that kills friendly soldiers is NOT against the laws of war. Generals don't get exemptions. I hope we see criticism of Russian behaviour at some point, rather than victim-blaming #Ukraine.

lol

Rest in pieces

Only elite VDVs get them.

INTRODUCTION Imagine an alternate universe in which people don’t have words for different forms of transportation—only the collective noun “vehicle.” They use that word to refer to cars, buses, bikes, spacecraft, and all other ways of getting from place A to place B. Conversations in this world are confusing. There are furious debates about whether or not vehicles are environmentally friendly, even though no one realizes that one side of the debate is talking about bikes and the other side is talking about trucks. There is a breakthrough in rocketry, but the media focuses on how vehicles have gotten faster—so people call their car dealer (oops, vehicle dealer) to ask when faster models will be available. Meanwhile, fraudsters have capitalized on the fact that consumers don’t know what to believe when it comes to vehicle technology, so scams are rampant in the vehicle sector. Now replace the word “vehicle” with “artificial intelligence,” and we have a pretty good description of the

AI Snake Oil is out today! The book has been five years in the making. The first chapter is online. It is 30 pages long and summarizes the book’s main arguments. We're grateful for all the early interest and support. Let us know what you think of the book! www.aisnakeoil.com/p/starting-r...

Yeah, it also depends if you're looking for a specific mutation (e.g., a drug target so you have to follow up with an orthogonal technology), or simply estimating a mutation count (e.g., tumor mutation burden).

I think there are 50 bins, so if you add up: height x (1/50), they should add up to one.

The actual BQs aren't represented in the right-hand figure. It just says in false positives, there is more of a difference in variant vs. reference-supporting reads.

And the fig on the right is like, if your BQs are 30, it's okay if it's 30 for both variant- and reference-supporting reads. But if it's 30 for variant-supporting reads but 38 for ref-supporting reads, then that is a problem.

Yeah, for a somatic mutation (much rarer than germline variants), you'd want the average BQs to be at around >=35 to be confident.

So many scientists in so many countries. Someone will do it if it can be done.

The right fig is interesting. For the TPs, the p-value histogram is flat, which is circumstantial evidence that the reported BQs are quite accurate relatively (i.e., higher BQ means more confident call). Whereas for the FPs, disproportion of them have different BQs for var- and ref-supporting calls.

Each of FPs and TPs add up to 1, so it's proportional of FPs and TPs with certain average BQs.
This is one of many features that boost the accuracy of the overall classifier, i.e., is a somatic mutation candidate an actual somatic mutation or not.

Two histograms showing likelihood of true positive vs. false positive somatic mutation calls based on base quality scores.

Here is an example. A xgboost model to call somatic mutations. BQs make up some features. The decision tree doesn't care 40 or 28, but it does care if a tree makes a cut in the middle.
The left is the average BQs supporting variant call. The right is a p-value between variant calls and ref calls.

Relative comparison is what I usually look for. Q40 and Q28 can be significant in a scenario where I look at two groups of reads supporting two different base calls, and help decide if one of them is a wrong call.