@slavov_n Yeah totally an important point. Nuclear transcriptome is shaped by export rates and has tons of pre-mRNA. Cytoplasmic by turnover and contains mostly mRNA. Great to show how different these are.
@disney_lab MYC is an interesting choice for this as its basal half-life in most contexts is very short (< 1 hour). Do you think MYC mRNA needs to be aberrantly stable for this particular approach to work well? Or is it really possible to significantly destabilize an already unstable mRNA?
Excited to announce our latest work! We present ARSENAL, a short-context DNA language model specifically designed to learn important sequence features in noncoding regulatory DNA.
Why is such a model important? Read on to find out!
https://t.co/3FyTSxxe5K
@genophoria This looks awesome. Small note: your SLAM-seq half-life estimation analysis is suboptimal. Median half-life in most human cell lines is around 4 hours, so inaccurate to assume that extent to which an RNA is labeled after 4 hours is entirely a function of transcriptional kinetics
@JMTali@IgorUlitsky That's very strange, but over the years our lab has had similar mysterious occurrences. The model we put forth in the prepint I linked above is that s4U RNA is sticky and can sometimes get lost on plastic surfaces during RNA extraction.
@IgorUlitsky Preprint: https://t.co/yHHJhKE2af
Paper: https://t.co/pN1FfK03xL
Seeing it strongly with a 30 minute label time is a bit surprising though. What s4U concentration and cell line are you using?
@IgorUlitsky Yes, there is a preprint and a paper on it. Preprint is from our (Matt Simon's lab) and paper is in NAR now and from Florian Erhards lab. Together these describe bioinformatic and RNA handling causes.
@em_maragkakis IMO Snakemake is a bit easier for biologists to pick up (combo of less abstraction and Python vs. Groovy). Nextflow's flexibility and the nf-core community are great though. I've liked teaching people in our lab how to make Snakemake pipelines and how to run nf-core ones.
@em_maragkakis Very exciting work! Do you think that 5EU provides unique advantages over other metabolic labels when it comes to reliably detecting it in direct RNA nanopore data? Or might your architecture easily generalize to detection of other labels?
@IgorUlitsky Supplementary Table 2 from the original SLAM-seq paper (https://t.co/FAffWr5uTV) is such a table from mES cells. It's 3-prime end sequencing data too, so should specifically fulfill the APA criterion.