slavov-n.bsky.social
Mentor, scientist & engineer. Having fun in @slavovlab.bsky.social and Parallel Squared Technology Institute @parallelsq.bsky.social with biology & single-cell proteomics.
https://nikolai.slavovlab.net
371 posts
2,124 followers
127 following
Prolific Poster
Active Commenter
comment in response to
post
Unbelievable. I hope this turns around soon.
comment in response to
post
The immediate cause for the excitement was new data from a recent improvement in a mass spec instrument that will be announced at #ASMS in June.
2/2
comment in response to
post
Expression noise facilitates the evolution of gene regulation
2/2
elifesciences.org/articles/5856
comment in response to
post
Deep Learning-coupled Proximity Proteomics to Deconvolve Kinase Signaling In Vivo
2/2
www.biorxiv.org/content/10.1...
comment in response to
post
Congrats
comment in response to
post
These results should contribute to making single-cell proteomics more robust, especially for applications to primary cells from tissues.
A good addition to the recommendations for best practices for single-cell proteomics:
single-cell.net/guidelines
www.nature.com/articles/s41...
comment in response to
post
The proteins exhibiting a significant change between permeable and intact cells were consistent across cell types and allowed us to build a generalizable classifier:
scp.slavovlab.net/QuantQC
comment in response to
post
This classifier is hosted in our single cell analysis package for samples prepared via nPOP (however the function just takes in any protein X cell matrix so can be easily applied to any data set) github.com/SlavovLab/Qu...
comment in response to
post
linkedin.com/posts/ralph-...
comment in response to
post
LLMs are impressive and powerful.
They can be fruitfully deployed in many areas, most obviously as augmenting Internet Search Engines.
They just do not have the capabilities to deliver on many of the extremely difficult tasks that AI evangelists advertise and proselytize.
comment in response to
post
AI evangelists posit that users are not good at "prompt engineering".
Yet, successful businesses rarely blame their customers for not being skilled or smart enough.
comment in response to
post
Even AlphaFold -- which is quite different and impressively successful -- has significantly lower performance with proteins dissimilar to the training data.
2/2
bsky.app/profile/slav...
comment in response to
post
Reading claims and being convinced of them are two different things. I have not seen convincing demonstrations of strong GPT performance on omics data resulting in nontrivial results that simpler models cannot achieve.
1/2
comment in response to
post
Compared to politics, false claims and premises in science and medicine have a lower chance of serving encounters with empirical evidence.
The concern of course is that such mistakes take rather heavy tolls before they are destroyed by the evidence.
comment in response to
post
... is designing experiments that minimize or even eliminate confounders when possible.
This requires measuring what is important, not just what is easy.
That's hard.
Understanding biology and developing effective therapies is hard.
Hype alone will not suffice.
2/2
comment in response to
post
Performance evaluation on RNA abundance reconstruction.
genomebiology.biomedcentral.com/articles/10....
comment in response to
post
An integrated landscape of mRNA and protein isoforms
2/2
www.biorxiv.org/content/10.1...
comment in response to
post
6 % indicated that the president should be able to ignore supreme court rulings.
This is 6 % too many.
comment in response to
post
This inverse correlation is observed across many proteoforms originating from alternative splicing, suggesting that proteoforms-specific stability (degradation rate) contributes significantly to establishing proteoform abundance, especially in the brain.
comment in response to
post
.. degradation rates substantially determine the proteoform abundance, especially in the brain.
Below is an example with Microtubule-actin cross-linking factor 1:
⬛️ The higher the degradation rate, the lower the abundance.