alevskaya

8 ms·

31.
by alevskaya 5y ago
Millions of old people in dense cities turn out to be a fantastically sensitive instrument for detecting the presence of sars-cov-2 in a population. You’d have to explain nearly two months of cryptic transmission that left no hospital repo
32.
by alevskaya 5y ago
I specialized in high throughput synthetic biology and in applications adapting vectors and pathogenic proteins to subvert the human immune system (for immunooncology mostly). So I have all the background needed for supervillain thinking h
33.
by alevskaya 5y ago
No one is arguing that the earliest known case on Dec 11 is when sars-cov-2 "began". We know it must be older than that. More recent reviews of chinese epidemiology don't mention the hubei patient, the Dec 11 case seems to
34.
by alevskaya 5y ago
Thanks - I don't expect to change anyone's mind. I just don't want to give into the cynicism that evidential reasoning is pointless. I don't want to tell anyone what to think... but given that I have the professional b
35.
by alevskaya 5y ago
Are you seriously posting a link to a one-off phylogenetic molecular clock estimate as experimental evidence of italian priority? The other Italian claims of priority are highly disputed by basically everyone - but have I missed anything -
36.
by alevskaya 5y ago
You're really into stacking these hypotheticals. Why would anyone bother developing a pile of fiddly new vectors? cui bono? Vector development sucks, I've done it in simpler virus families. And why do it completely markerless?
37.
by alevskaya 5y ago
The trouble with the synthetic furin hypothesis and all this DRASTIC stuff on that grant, etc. is that all of this work would certainly be done in known viral vectors... you would never go messing around testing spike proteins in some total
38.
by alevskaya 5y ago
...but that report doesn't contradict anything. The current earliest known case is a Huanan market seafood vendor who contracted the disease on Dec -11-. Dec 24 is way late in the existing epidemiological timeline. (And no one is sa
39.
by alevskaya 5y ago
Per the CCDC 33 of 585 environmental samples taken early Jan. were positive, and 31 of those were concentrated near the live-animal region of the Huanan market. Environmental PCRs aren't magic, they have a finite level-of-detection, R
40.
by alevskaya 5y ago
An epidemiological nexus is much smaller than an entire city and the details of the early cases matter. Aside from the contact-tracing, investigators were able to get positive PCR recovery of sars-cov-2 from -environmental samples- at the
41.
by alevskaya 5y ago
CS people always underestimate the difficulty of this stuff. I used to be a genetic engineer: Even in the best of times doing real field science on the lineage analysis of sars-cov-2 would take many many years. In a time of acute politic
42.
by alevskaya 5y ago
That first call is from tf2xla which JAX doesn't use. But I think those calls just set up the global buffers used by the launched program inputs/temporaries. The entire point of XLA's static shape discipline is allowing for
43.
by alevskaya 5y ago
One thing that's important to note about memory management with XLA, is that inside a compiled program there's no user-exposed "malloc"/"free". The memory usage and schedule of a given program/graph
44.
by alevskaya 5y ago
Buddy, this is human medicine, you don't get the luxury of "hard numbers" as if there was an SQL table in the sky you could just query. There is the CDC's VAERS reporting system, but it doesn't define reality: When
45.
by alevskaya 5y ago
This is a terrible, uninformed take. Had we seen any serious unwanted side-effects during the EUA authorizations the FDA would withdraw the authorization, and the vaccines have continued to have excellent track records across over a hundre
46.
by alevskaya 5y ago
Proving the zoonotic chain of transmission is not a quick or easy thing to do even in a cooperative geopolitical environment. This is not a field that lends itself to snappy "smoking guns". It took years to build the evidence fo
47.
by alevskaya 5y ago
KEGG is a very useful index into known genes for this, I used it all the time when mining for possible enzymes in the past.
48.
by alevskaya 5y ago
The paper you link to is overexpressing LINE1 on plasmid in HEK293 cells and coinfecting w. sars-cov-2 - not exactly a natural setup! They see some nucleocapsid fragment integration and limited expression, but it's a pretty forced exp
49.
by alevskaya 5y ago
In all of your comments here you're being pretty ridiculous - you've read some books and articles, but are painting stories here in such a credulous fashion that makes me think you haven't actually done much work with cells a
50.
by alevskaya 6y ago
This isn't correct. The constraints placed upon epitopes by the cell receptor vs antibodies aren't the same. Rhinoviruses and Influenza viruses escape past immunity without fundamentally altering their biology all the time! Th
51.
by alevskaya 6y ago
Averaged over a population - some people will have resistant sera and others not depending on the parts of sars-cov-2 targeted by their polyvalent antibody response. Note that this is only looking at antibodies, not T-cell responses... so
52.
by alevskaya 6y ago
The first study is only studying the effect of the N501Y mutation as found in several emergent strains, not the E484K mutation.
53.
by alevskaya 6y ago
It's unlikely to _completely_ escape immunity - humans will raise antibodies against different regions of the spike protein, but it does seem to confer a pretty significant average resistance to patient sera. There's another stud
54.
by alevskaya 6y ago
What I meant by sera above was general adaptive immunity. It's likely to partially resist past immunity from both vaccines (esp. from the spike-only vaccines: moderna, pfizer, etc) and natural infections from older strains. We'l
55.
by alevskaya 6y ago
Sadly, note that the study talked about here [1] is investigating the effect of the N501Y mutation, not the more worrying E484K mutation found in the South African 501.v2 strain that likely does escape antibody drugs and reduces neutralizat
56.
by alevskaya 6y ago
The Dex authors are also core Autograd & JAX authors, so they're not unaware of the benefits of the dense Numpy-style APIs. The challenge is that many numerical kernels and algorithms are just not easily or naturally expressed in
57.
by alevskaya 6y ago
JAX and Pytorch have somewhat different scopes though... JAX is concerned about much more than the kinds of neural nets we write today. It's a general system for expressing and transforming numerical programs, and the devs are as genu
58.
by alevskaya 6y ago
Having written and debugged both Theano and TF plenty in the past, I think this is a somewhat uncharitable take, esp. recalling the absolutely enormous Theano compile times. :) I think Theano was genius, but a system that relied on python-
59.
by alevskaya 6y ago
JAX autograd will work on most any jitted fn - the control-flow limitations are no autograd for code with for/while loops since there's a statically unknowable trip count through the loop body. Much looping code can be handled di
60.
by alevskaya 6y ago
On this topic I highly recommend the book "Libyan Sands: Travel in a Dead World" by Ralph Bagnold, an account from a British officer who led some of the first deep Sahara expeditions with Model Ts! Lots of great stories of gettin

More ›