"Deep thermal profiling for detection of functional proteoform groups" H/T: (.@savitski_lab)

"Ultra-high throughput mapping of genetic design space" H/T: (.@BashorLab)

"Precise programming of multigene expression stoichiometry in mammalian cells by a modular and programmable transcriptional system" H/T: (.@YangLiu_LMB)

"Predicting EGFR mutational status from pathology images using a real-world dataset" H/T: (.@BostonGenomics)

Here's what I'm reading so far this week. Thanks very much to the original posters (tagged below): "Protein Structure Prediction with In-Cell Photo-Crosslinking Mass Spectrometry and Deep Learning" H/T: (.@slavov_n)

@Biotech2k1 @RecursionChris For germline applications, there will be a lot of mosaicism and for cancer, you’d need access to fresh-frozen tissue—so there are certainly obstacles. That said, I like the idea of digging deeper into alternative splicing and diseases caused by dysregulation therein.

@Biotech2k1 @RecursionChris It’s interesting you should ask! Chris and I recently talked about this. I won’t speak for him, but from my point of view, this is an upcoming wave—labs that operationalize high-throughput long-read techniques (e.g., MAS-Seq) will be in the pole position.

@Biotech2k1 @TohJah7809

@Biotech2k1 @TohJah7809 Yes, my understanding is VHH nanobodies have unique advantages insofar as they're smaller and able to penetrate more deeply into solid tumors. Alpacas produce high quantities of VHH when immunized, making screening more efficient as well.

A terrific review of cutting edge techniques for sequencing peptides at the single-molecule level—by .@metricausa.

@AlbertVilella @Mr100years @OmicsOmicsBlog @Phil_A_Richmond I'm not quite sure. Regardless, it spreads out the discussion to a bunch of separate threads, makes it difficult to bookmark and follow, and if I quote-Retweet, then exposes an esoteric discussion that I'm engaged in to many people who don't care (even though I do).

@AlbertVilella @Mr100years @OmicsOmicsBlog @Phil_A_Richmond No, it's very clunky and discourages free discussion/debate (IMHO).

@AlbertVilella Please also collapse duplex yield if you wouldn't mind, it's still being double-counted. ^ Separately, do you have any data on what short-read companies are offering for bulk discounting?

@AlbertVilella Assuming $600/FC, 85% utilization, & 72-hour run time, I get ~2,482 FCs / year for P24. With negligible compute and $225K CapEx, I get to ~1,034 30x WGS (~90 GB) at $16/GBp for P24.

@AlbertVilella A few questions: Why do some platforms get bulk reagent discounts while others do not? The unique duplex yield / Prom FC is ~30 GBp, not 60 Gbp? In my calc, I assume 25% unique duplex on 150GBp total yield, so ~37.5 Gbp / FC. 1/2

Retweeting for visibility, thanks for announcing this during this afternoon’s #AGBT23 session, .@coregenomics!

@MarceloPLima …sequence in ‘real-time’, unlike short read systems. To my knowledge, these teams are totally separate. I don’t know for sure, though.

@MarceloPLima Yes, they’re very talented bioinformaticians and deep learning experts. DeepConsensus uses a transformer-encoder architecture with principles adapted from NLP—it’s also tuned to mitigate insertion-deletion (indel) errors that plague long-read systems since they both…

@MarceloPLima DeepVariant is compatible with most flavors of sequence reads and is a tremendous variant caller. To your point, though, DeepConsensus lives upstream of variant calling and, to my knowledge, is the only one of these tools that runs on the device itself to generate base calls.