and btw, here is a much better write-up by @pwilmarth.bsky.social from 2021. pwilmart.github.io/blog/2021/04... I just did my own internal write-up and leaned on this completely.

I'm amazed at academics using AI. You are whitewashing the AI field for tech bros and they have rewarded your efforts with funding cuts, job losses, and incredible stress. But go ahead and keep writing those AI papers and having AI conferences. Maybe you'll raise your h-index before you're erased.

Size matters. How big is your embedding? Maybe we should call AI what it really is: Alpha-male Insecurity. If your research work involves AI, just stop. It is like saying you found a couple of non-mass-murder uses for assault rifles, so you recommend all researchers start using them.

ECD is back! That took about 20 years. Proteomics must be slowing down. The length for proteomic idea recycling is more like 10 years. Maybe ECD results in longer time distributions? Since it is shooting electrons at big things, I hope it uses a manual joystick. Turn MS into a mass player e-game!

I can confidently predict that "inflammation" and "immune response" will be associated with many diseases...

When you are mostly concerned about the size of your study, can we all agree to call that a search for bromarkers?

Windows (10 or 11) is total shit thanks to AI and MS being the World's worst company. Win10 virtual machine makes my iMac fans run all the time. Pausing Win10 stops the energy wasting. Resuming Win10 results in *several* unusable minutes of 100% CPU use as Win10 tries to catch up on spying. FU MS.

Ha-ha. Proteomics methods from 15 years ago are not ready for prime-time. How many wrongs (bad sample prep, bad enrichment, bad LC, bad MS, bad data acquisition, bad data analysis, bad experimental design, bad QC) does it take to make a right? We need more AI! AI beads, AI LC, AI MS, AI marketing...

If your issue with Cracker Barrel is the logo, then you must have your own portable defibrillator...

We are lucky that proteins are robust molecules. No spontaneous peptide bond breaking, no artifactual modifications from oxygen or buffers, and they never stick to any surfaces! Why were we ever so stupid to digest with trypsin and try (with limited success) to use mass specs to sequence peptides?

I'm co-author on a paper that was submitted to MCP. It was rejected because it was not *novel* enough. It used thoroughly tested quantitative proteomic techniques and had reliable, believable results. I think that makes the paper *extremely* novel for MCP (or any other proteomics journal).

I wanted to get a couple more years of work in before retiring. Was that too much to ask? Now, there's funding uncertainty, and GitHub is being forced to all AI crap by Microsoft. Is there nothing that MS can't turn to shit? I will be looking for a GitHub alternative and wasting time migrating...

Do you need an AI's help to write a blank review???

If this was for microscope data, they could have called it CryptoEM...

It is the 100th anniversary of the Pig -n-Ford races this year: www.tillamookheadlightherald.com/news/pig-n-f... The racing starts today and the World Champion will be crowned on Saturday! My money is on Wout van Aert...

A really dumb question that just won't die is: "when are they going to make the women's Tour de France more like the men's race?" The question should be when are they going to make the men's race more like the women's race. This year's women's tour is absolutely amazingly great. OMG!!!

The $10 proteome is interesting. Assuming 100 nanograms of digest per sample to get a really shitty result, that is $100 million per gram. There are 453 grams per pound, so that is 45.3 billion dollars a pound for manure. I should have stayed on the dairy farm. Talk about cash cows!

It looks like Sydney Sweeney's "great jeans" ads might be a big bust...

I fixed the title: "Non-random extracted ion chromatogram peak areas as a novel complex, computationally-heavy data imputation method for non-quantifiable proteins in quantitative proteomics: imaginary data solves the complex number proteomic crisis in experiments with negative replicate numbers."

Finally, the real Tour de France starts today. This 3-week Tadej de France snooze fest is so over. If they had a mercy rule in grand tours, it would have ended two seeks ago. Now it is time for some real bike racing.

The NIH NCI announces that cancer research funding is in near total remission...

Didn't get enough #TeamMassSpec at the Cascadia Proteomics conference last week? Did our @talusbio.bsky.social presentations pique your curiosity? Say no more because Bruker has you covered. I'm speaking at the "Bringing ASMS 2025 to Seattle" event TODAY! Registration still open! lnkd.in/gTs62ePq

I can't believe how many attractive young women with cute user names are really into proteomics! You know it is a healthy social media platform when you have such a diversity of followers!. Do the women scientists get a lot of proteomics-loving hunky young men with rugged user names as followers?

I wonder how UAE and Tadej will attack his rivals on the rest day. Can he win the rest day? How much time can he put into his rivals? Did the team recon the rest day? What is the best rest day bike setup? What time bonuses and points are on offer for the rest day?

After 22 years of proteomics data analyses, my toolkit of software has expanded to Comet, edgeR, BLAST (locally installed version), and Excel. I could include Jupyter notebooks and some R (ggplot2), I suppose. Programming tip: avoid software dependencies. Science tip: avoid software dependencies.

I was going to post the Epstein client list, but the deep state stopped me. You'll have to settle for thoughts on how to prep/filter quantitative proteomics data before doing QC and statistics (github.com/pwilmart/qua...). You can't really understand the experiment if you don't understand the data.

Some replies to my post about detecting deamidation (github.com/pwilmart/det...) mentioned that RT predictions could help. While deamidation looks like an amino acid substitution, it is more complicated. See this just-accepted article (paywalled) for more details: pubmed.ncbi.nlm.nih.gov/40654079/

Proteomics 101: detecting deamidation is easy - just specify a variable modification in your favorite search engine, right? We have been using search engines to detect modified peptides since 1995, so it must be all worked out by now... github.com/pwilmart/det... (work in progress, may change)

Proteomics data analysis has been so bad for so long, we have ended up in strange places. Folks are working on wholesale data fabrication, which is scientific fraud, and calling it data imputation. Single-shot quantitative experimental designs are one of my biggest gripes: github.com/pwilmart/PXD...

A friendly reminder that the best stage race in July started today: www.giroditaliawomen.it/en/ Real racing, real riders, and a lot less circus.

DIA, DOA, DUI, DDA, etc. Here is a comparisons of some quantitative proteomics methods from a POV you might not have seen before: github.com/pwilmart/qua...