Hacker News new | past | comments | ask | show | jobs | submit login

Things are pretty different from what I was taught in high school now, or perhaps just more detailed. Some examples:

Protein and ligand are like a lock and a key >>> Some proteins have no intrinsic shape and wrap around ligands or other proteins. Maybe there is RNA involved as well.

The cell membrane is like a soap bubble >>> Different lipids cluster together to form raft like structures with different content based on lipid composition.

DNA -> RNA -> Protein >>> DNA <-> RNA -> Protein.

We have a lot of Junk DNA >>> We shouldn't call it junk DNA, it plays a role in regulation of the exome, wrapping DNA around chromatin and probably plays a lot more roles on a larger scale, equally or more important than the exome even.

Optical microscopes have resolution a limit of about lambda[the wavelength of the used light]/2 >>> We have super-resolution optical microscopy now (ie STED, STORM/PALM, SMM probably more and better, I left the field 10 years ago).

The immune system plays no role in cancer >>> We're completely removing metastatic cancers by boosting the immune system regularly now.

Just to make you think, when I was 19 I did Sanger sequencing, about a day or 2 of hands-on work for 200 base pairs. Now I'm 39, we have an Illumina Nextseq 500 in our lab, we're sequencing 1200 billion (1.2e11) base pairs in one experiment (assuming 2x150 bp reads, 400 million reads). A Novaseq can do a lot more... Crazy.




Is sequencing 'hard' computational work? Could we use it for proof-of-work?


Aligning all the short reads a (Sequencing By Synthesis) Next Generation Sequencer (to be very specific) produces (hundreds to thousands of millions) to a reference genome is hard computational work indeed, no idea if you can use it for proof-of-work but I like where this is going... But do you not need some kind of definition of correct (like hash needs many leading 0's) for PoW? How would that work for aligning reads to a reference genome? Maybe it could, find the position of the read is hard (I mean, not really for a modern CPU but relatively... I think... not an expert on that), verifying how correct it is, is not hard (I think). Maybe we should ask the BWA devs [0].

[0]: https://github.com/lh3/bwa


I’d be in line to purchase some HengCoin. It’s an interesting idea but you also raise a good point that alignment is probabilistic.

It is the step that most benefits from cpu parallelization so would be cool to explore at least


What do you mean by alignment being probabilistic? If you mean that some reads align faster and you could game the system, then perhaps we should tie the system to hospital based verification of the input data.

Ie, hospital publishes sequencing data (perhaps using fully homomorphic encryption to keep this very sensitive data private), someone aligns it, someone (or the hospital) verifies it, aligner gets paid coin.

The verification of the alignment being easier than the alignment itself is analogous to blockchain verification being easier than actually signing a block, right?


Not quite the best phrase I'll admit. It's identifying the best fit. That's not the same as identifying truth. For instance, my recollection is that bwa-mem has (had?) some small amount of nondeterminism in it.


Ah yes, it would be easy to confirm a good alignment, identifying if any alignment is the best there is would be more difficult I guess, indeed.




Consider applying for YC's W25 batch! Applications are open till Nov 12.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: