Paper Highlight: Beyond MS, The Next Step in Proteomics

I’ve been posting a lot to keep myself reading and engage all of you! Please comment below!

Here’s the paper.

Why proteomics?

• Proteins are important! Did you know Sanger studied protein sequence first?
• Sequencing is awesome (and cheap), but it isn’t enough! Transcription levels are easy to measure (RNA-seq), but don’t necessarily correlate to protein levels (translation kinetics, RNA modifications, PTMs, degradation, etc.).
• Sequencing doesn’t even necessarily tell you which protein it is! Short reads don’t deal well with isoforms and long reads have indel errors that can cause frameshift misinterpretations.

How do we use MS for proteomics?

• “Bottom up”: snip the proteins into smaller pieces, put them through a separation, and then match the fragments against a database. Note that this doesn’t reall sequence the protein directly (like Edman degradation would), but rather tries to find protein fingerprints and infer how many copies there are.
• Bottom up sensitivity is limited by S/N in the MS. <1% of ions are actually used because most fragments are so low in count! Sensitivity also affects dynamic range. Orbitrap MS has a range of 1E5 and that’s not nearly enough to cover the 1E12 range in a clinical specimen. You need >1E6-1E9 copies to see a protein. Even then, you can be confounded by isoforms and PTMs (like phosphorylation!).
• “Top down”: no snipping, just detect intact proteins with ultra-high-resolution FTMS. Super-high-res is needed because resolving PTMs requires <1 ppm accuracy, which gets harder when you have a heavier protein. You get isoforms and PTMs, but it’s 100x less sensitive and it’s hard to get intact protein ions for proteins >70 kDa.

What’s the dream?

• Sequence proteins using nanopores, which are currently used for long-read DNA sequencing. You embed a pore into a thin membrane, apply a voltage bias, and then thread unwound protein through the pore. The current depends on the sequence, PTMs and all, and you get single molecule sensitivity!
• Unfortunately, there aren’t suitable pores yet. The pores for DNA are way too big to distinguish between AAs. You also need a smaller pore to slow translocation through the pore.
• We also don’t have suitable membranes, either. The protein needs to be denatured, and denaturants will wreck the lipid membranes that are currently used for DNA sequencing.
• Ideally, translocation would be uniform and unidirectional, but proteins aren’t uniformly charged.
• The authors are working on tiny inorganic pores and membranes that will be small enough to read individual AAs and robust against denaturants. They’re proposing ML algorithms to deal with velocity and read errors (sadly common still).

It’ll be super cool if works! I wonder if we could store ultra-high-density information in custom polymers and read them out with pores? Wouldn’t it be cool if you could slip the entire internet into a test tube and preserve it for the deep future?