Portal Protein Pattern Pinpointing Progress
A few folks have sent me this article from GenomeWeb on recent progress in protein sequencing using (protein) nanopores. This focuses on the work of Portal Biotech, a London/Netherlands based nanopore proteomics startup.
I know a number of folks there, they’re all awesome, though I have no specific information about the progress they’ve made at Portal.
So let’s take a look at some of the statements made by Portal in this article:
Heron said the company "routinely measures proteins that are more than 100 kilodaltons in size.” He noted that one advantage of measuring such large proteins, as opposed to shorter peptides, is that their signature is highly unique, making for easier identification.
This makes sense to me! Comparing one large protein to another very different large protein seems like the easier challenge. As I understand it 100 kilodaltons is going to be in the 1000 amino acid range.
We can get some sense of how easy/difficult this might be by looking at the number of possible states (20^1000). Google helpful tells me that 20 to the power 1000 is… infinity:
While this is obviously incorrect, it’s certainly a very big number… I suspect comparing two very different 100 kilodalton proteins would be “easy”1, two very similar ones… very hard.
But as a place to start it seem like a very solid idea. Bring up the platform and generate reliable distinct signal from large proteins. Then I imagine optimize and iterate.
Heron said the company is now at the point where it is capable of producing such simulated databases and identifying individual proteins out of thousands of potential candidates with accuracies above 90 percent, though it has not yet published peer-reviewed research detailing this work.
90% accuracy seems pretty good. Devil is in the details of course. What exactly is the experiment. What are the potential candidates (is this the database generated from simulation mentioned?). I guess we’re not talking single AA differences here, or PTMs?
How do you measure accuracy? A known sample which contains only a limited number of proteins? Or are we looking at a more complex background? Or is this quantification of the relative protein content of a more complex sample.
Many proteomics applications seem more like counting. So, perhaps as I previously suggested you can group similar proteins, then generate a consensus trace that helps you identify smaller differences (like PTMs)?
According to Heron, the company, which raised $11 million in venture funding in 2022, has begun placing instruments with early adopters ranging from academic labs to small and large biopharma companies, though he declined to name any specific firms. The company is also analyzing samples sent by outside customers, Heron said. "It's still at the stage where we continue to refine the software and capabilities, but the most important thing is for us to put it in the hands of our customers, for them to tell us how we can improve it further for their particular applications.”
Bringing up a functional nanopore platform is an expensive exercise. I suspect $11M doesn’t buy you much in terms of building a practical commercial instrument. But who knows, Portal may have found a neat hack to avoid costly (and slow) ASIC design and array fabrication projects commonly associated with nanopore platform development.
In any case, it’s starting to feel like we will have practical nanopore based proteomics platforms. The question is more if these will remain niche research tools or gain wider adoption.
Will be interesting to see how things develop!
For definitions of easy which include the huge amount of effort required to get any Nanopore detection platform up and running. And of course, these is predicated on solving the issue of getting proteins to translocate at all and at a somewhat detectable rate.