Jase Gehring
@skyjase.bsky.social
scientist at UC Berkeley inventing advanced genomic technologies
lover of molecules, user of computers
https://scholar.google.com/citations?user=63ZRebIAAAAJ&hl=en
lover of molecules, user of computers
https://scholar.google.com/citations?user=63ZRebIAAAAJ&hl=en
big opening for an Epstein files live coding sesh
November 13, 2025 at 5:18 AM
big opening for an Epstein files live coding sesh
mesmerizing wagon-wheel effect using hi-speed camera and droplet generator. droplets enter the channel from the top, flowing counterclockwise around the bend, a mixing feature. droplet generation rate is ~1kHz, droplet size ~80 microns
September 30, 2025 at 7:55 PM
mesmerizing wagon-wheel effect using hi-speed camera and droplet generator. droplets enter the channel from the top, flowing counterclockwise around the bend, a mixing feature. droplet generation rate is ~1kHz, droplet size ~80 microns
funky bubbles not for human consumption
September 23, 2025 at 3:50 PM
funky bubbles not for human consumption
sequence diversity seems high. no blast hits and foldseek sequence alignments show low similarity.
colabfold pLDDT is high, but there's big disagreement on the N terminus there. against the foldseek hit, also no alignment on the N-terminus.
colabfold pLDDT is high, but there's big disagreement on the N terminus there. against the foldseek hit, also no alignment on the N-terminus.
September 23, 2025 at 1:32 AM
sequence diversity seems high. no blast hits and foldseek sequence alignments show low similarity.
colabfold pLDDT is high, but there's big disagreement on the N terminus there. against the foldseek hit, also no alignment on the N-terminus.
colabfold pLDDT is high, but there's big disagreement on the N terminus there. against the foldseek hit, also no alignment on the N-terminus.
checking out my first samples from la-proteina
this is a 200-mer inferred with unconditional generation using their baseline v1.1 model. la-proteina structure in blue. L: alignment with colabfold inference. R: partial alignment w foldseek hit (rmsd 8.2A)
research.nvidia.com/labs/genair/...
this is a 200-mer inferred with unconditional generation using their baseline v1.1 model. la-proteina structure in blue. L: alignment with colabfold inference. R: partial alignment w foldseek hit (rmsd 8.2A)
research.nvidia.com/labs/genair/...
September 23, 2025 at 1:32 AM
checking out my first samples from la-proteina
this is a 200-mer inferred with unconditional generation using their baseline v1.1 model. la-proteina structure in blue. L: alignment with colabfold inference. R: partial alignment w foldseek hit (rmsd 8.2A)
research.nvidia.com/labs/genair/...
this is a 200-mer inferred with unconditional generation using their baseline v1.1 model. la-proteina structure in blue. L: alignment with colabfold inference. R: partial alignment w foldseek hit (rmsd 8.2A)
research.nvidia.com/labs/genair/...
my first #FluorescenceFriday
here is a solution of gelatin. purchased from Sigma, resuspended in PBS, under blue LED illumination.
🤨 idk #WhatsTheMechanism
here is a solution of gelatin. purchased from Sigma, resuspended in PBS, under blue LED illumination.
🤨 idk #WhatsTheMechanism
September 19, 2025 at 5:20 PM
my first #FluorescenceFriday
here is a solution of gelatin. purchased from Sigma, resuspended in PBS, under blue LED illumination.
🤨 idk #WhatsTheMechanism
here is a solution of gelatin. purchased from Sigma, resuspended in PBS, under blue LED illumination.
🤨 idk #WhatsTheMechanism
this is the most important plot in the paper. specifically the two sub-sub-panels that i'm highlighting. the one on the left is a histogram of 'average amino acid identities' and the one of the right is a histogram of the similarity of the spike protein to the parent PhiX genome.
September 18, 2025 at 3:30 AM
this is the most important plot in the paper. specifically the two sub-sub-panels that i'm highlighting. the one on the left is a histogram of 'average amino acid identities' and the one of the right is a histogram of the similarity of the spike protein to the parent PhiX genome.
the breadth and detail, the weaving of history, historical data, and the scientists who made the breakthroughs.
"this book is like the Bible, but you'll learn a lot more about Life"
"this book is like the Bible, but you'll learn a lot more about Life"
September 16, 2025 at 1:00 AM
the breadth and detail, the weaving of history, historical data, and the scientists who made the breakthroughs.
"this book is like the Bible, but you'll learn a lot more about Life"
"this book is like the Bible, but you'll learn a lot more about Life"
it's spittin bullshit lmao. you can just say things
September 11, 2025 at 9:01 AM
it's spittin bullshit lmao. you can just say things
nice work! so i'm like is that just a Swedish-sounding acronym and they went with it, threw in an umlaut for fun? (in which case, cheers)
or is it a real Swedish word/backcronym?
and i can't tell if google is having a stroke or just fucking with me
or is it a real Swedish word/backcronym?
and i can't tell if google is having a stroke or just fucking with me
September 11, 2025 at 9:01 AM
nice work! so i'm like is that just a Swedish-sounding acronym and they went with it, threw in an umlaut for fun? (in which case, cheers)
or is it a real Swedish word/backcronym?
and i can't tell if google is having a stroke or just fucking with me
or is it a real Swedish word/backcronym?
and i can't tell if google is having a stroke or just fucking with me
SPCs are made by mixing two polymer solutions using a microfluidic droplet generator. Under the right polymer concentrations, they will phase separate (ATPS) into a core and a shell with contrasting polymer ratios. (my photo below).
September 10, 2025 at 4:36 PM
SPCs are made by mixing two polymer solutions using a microfluidic droplet generator. Under the right polymer concentrations, they will phase separate (ATPS) into a core and a shell with contrasting polymer ratios. (my photo below).
if you haven't heard of semi-permeable capsules (SPCs), keep an eye on this emerging single cell omics tech!
SPCs have a thin hydrogel shell surrounding a liquid core. they're miniature reaction vessels that can trap cells and large molecules while allowing nutrient, waste, and reagents exchange 🧪🧵
SPCs have a thin hydrogel shell surrounding a liquid core. they're miniature reaction vessels that can trap cells and large molecules while allowing nutrient, waste, and reagents exchange 🧪🧵
September 10, 2025 at 4:36 PM
if you haven't heard of semi-permeable capsules (SPCs), keep an eye on this emerging single cell omics tech!
SPCs have a thin hydrogel shell surrounding a liquid core. they're miniature reaction vessels that can trap cells and large molecules while allowing nutrient, waste, and reagents exchange 🧪🧵
SPCs have a thin hydrogel shell surrounding a liquid core. they're miniature reaction vessels that can trap cells and large molecules while allowing nutrient, waste, and reagents exchange 🧪🧵
Some droplets for y’all, as a treat 🫧
September 8, 2025 at 5:20 PM
Some droplets for y’all, as a treat 🫧
this article starts with an illuminating misunderstanding. Base LLMs will totally spit out streams of nonsense without a prompt. as with LLMs, GLMs won't be very useful without extensive post-training, and even then they'll make tons of mistakes. we are severely data limited in this domain
August 19, 2025 at 7:27 PM
this article starts with an illuminating misunderstanding. Base LLMs will totally spit out streams of nonsense without a prompt. as with LLMs, GLMs won't be very useful without extensive post-training, and even then they'll make tons of mistakes. we are severely data limited in this domain
August 7, 2025 at 3:11 AM
IMO it's not crazy, but this kind of deep digging isn't just expensive/inefficient, it's way more prone to technical artifacts and requires a lot of skepticism and scrutiny from the researchers. per base error rate in this kind of library is probably >1%
June 27, 2025 at 2:56 PM
IMO it's not crazy, but this kind of deep digging isn't just expensive/inefficient, it's way more prone to technical artifacts and requires a lot of skepticism and scrutiny from the researchers. per base error rate in this kind of library is probably >1%
fig. 1b has their distributions, and i'm not sure offhand how it compares to previous datasets, but it's plausible. you see the end bias.
their table has reads per cell but not UMIs per cell, which IMO is an important comparison that is left out. further rarefaction analysis is warranted
their table has reads per cell but not UMIs per cell, which IMO is an important comparison that is left out. further rarefaction analysis is warranted
June 27, 2025 at 2:56 PM
fig. 1b has their distributions, and i'm not sure offhand how it compares to previous datasets, but it's plausible. you see the end bias.
their table has reads per cell but not UMIs per cell, which IMO is an important comparison that is left out. further rarefaction analysis is warranted
their table has reads per cell but not UMIs per cell, which IMO is an important comparison that is left out. further rarefaction analysis is warranted
more from figure 5. what would success and failure look like in this experiment? at best we are seeing vaguely zebrafish looking 'gross morphology'
May 10, 2025 at 11:42 AM
more from figure 5. what would success and failure look like in this experiment? at best we are seeing vaguely zebrafish looking 'gross morphology'
the 2025 paper similarly lacks clear confirmation between imaging and DNA reconstruction. Figure 5, by my reading, shows 3D inference on the right for zebrafish specimens on the left. again, they don't look similar! what is the reader supposed to glean from these data?
May 10, 2025 at 11:42 AM
the 2025 paper similarly lacks clear confirmation between imaging and DNA reconstruction. Figure 5, by my reading, shows 3D inference on the right for zebrafish specimens on the left. again, they don't look similar! what is the reader supposed to glean from these data?
this is from figure 4 of the 2019 DNA microscopy paper, purporting to show 'accurate reconstruction of fluorescence microscopy'
if i understand correctly, the bottom reconstruction should look a lot like the top image, and to me it really doesn't! so, did it work?
if i understand correctly, the bottom reconstruction should look a lot like the top image, and to me it really doesn't! so, did it work?
May 10, 2025 at 11:42 AM
this is from figure 4 of the 2019 DNA microscopy paper, purporting to show 'accurate reconstruction of fluorescence microscopy'
if i understand correctly, the bottom reconstruction should look a lot like the top image, and to me it really doesn't! so, did it work?
if i understand correctly, the bottom reconstruction should look a lot like the top image, and to me it really doesn't! so, did it work?
except big surveillance tech. they are compliant and necessary
photo credit: Google Maps
photo credit: Google Maps
April 9, 2025 at 10:32 PM
except big surveillance tech. they are compliant and necessary
photo credit: Google Maps
photo credit: Google Maps