On this episode of Translating Proteomics, Parag and Andreas share their reflections on proteomics developments in 2025 largely inspired by their observations at the World HUPO 2025 conference in Toronto. Whether you agree, disagree, or simply want to share some of your own observations on proteomics, please post your thoughts in the comments.

We look forward to even more exciting developments in 2026!

Chapters

00:00 - 00:35 – Intro

00:36 – 07:00 - Increased focus on applications of proteomics and less on method development

Learn more about One Health from our conversation with Professor Jennifer Geddes-McAlister

https://youtu.be/JFwvTdkb5bw

07:01 – 12:47 - Increase in people talking about the importance of proteoforms

Learn more about proteoforms in our episode featuring proteoform pioneer Professor Neil Kelleher

https://youtu.be/3pPuxVrMxpw

12:47 – 17:26 - An increase in multiomics studies

17:27 – 20:03 - A shift to larger scale proteomics studies

For a great example of a multi-platform comparison study, check out Kirsher et al., 2025

https://www.nature.com/articles/s42004-025-01665-1

20:03 – 25:07 - Increased integration of AI into proteomics workflows

For an example of how proteomics workflows can be modified with multiomic data, check out Suhre et al., 2025

https://www.nature.com/articles/s41588-025-02413-w

25:08 – 30:05 – Recognition of the need to assess quality across proteomics workflows

30:06 – 32:19 – Less of a focus on spatial proteomics this year than in past years

32:20 – 35: 13 - Parag and Andreas share their predictions for 2026

35:14 – End – Outro

Foreign.

Welcome back to Translating Proteomics.

And whether you agree or disagree with anything you hear in this episode, we'd love to hear from you.

So please share your thoughts, your reflections from the year in the comments and we may feature them in a future episode.

All right, Andreas, ready to dive in?

Yes.

Let's get started.

Great.

Well, so maybe I'll start with you.

What are for, for the year, what was your largest, largest takeaway either from world Hoopoe or from the field of proteomics overall?

Yeah, I mean the, the, the, the general observations throughout the year and what was happening in Hoopoe, I think are converging.

From my perspective, there was essentially a visible, much fewer posters and presentations in Toronto than maybe in previous year.

But the posters and presentations we saw were much sharper in focus, more meaningful, and it was very much application driven research.

The particular field that stood out for me was that over 25% of the posters were essentially functional translational proteomics applications.

And within that series of presentations, the clinical proteomics took center stage.

And within that category of clinical proteomics, immunopeptidomics was basically one of the main areas of highlights at hopo.

And maybe we describe what immunopeptidomics is, but in general, a cell generates essentially small peptides that it presents on the major histocompatibility complex on the surface of the cell, and then the immune system essentially learns what's present in the cell or tissue.

These particular peptides that are being presented are particularly important in the context of oncology, autoimmune diseases, but also in a precision medicine approach.

So having the ability to understand what the immune system essentially sees is a core application of proteomics can't be done by any other technology.

And it's actually a very, very nice sweet spot for high resolution mass spectrometry.

And so that was very, very clear for a front runner in applications at hopo.

Did you see that?

That's interesting.

So I mean, just to double click a little bit, it sounds like you saw overall a slight shift in the field from the percentage of posters that were purely technology focused to ones that were more focused on the application.

Would you say that it is a larger percentage of folks who are using proteomics as a tool to answer other biological questions?

Do you think those questions are further along now so that they have more to share.

What do you attribute this change in this subtle shift in the field to?

That's a great question.

I think over a decade and a half we have seen mostly focus around technology.

How do we evolve our methods development?

And that basically is now no longer the main focus for hupo.

There's about 10 to 15% of the posters.

We will see new method development.

People demonstrate that they can do that in a new context.

But there's much more focus on true application of the technologies to biological questions.

And so that's very exciting because the field itself now has, amongst many of the contributions, real insights, real biological insights.

That's interesting.

I'm not sure I observed that.

I think I definitely saw a widening.

That there were.

There were a larger number of more biology focused studies being shared.

But I think my takeaway was that the field still was very much in a development comparison mode.

There were a large number of studies that were highlighted about comparing this method to that method.

And so my take was that while there was this shift and it was driven by people who were really super users of the technology and were working very closely with biologists and clinicians, I agree with you that that was maybe about 25% of what.

What was there.

And then they're probably, I think in your characterization, about 15% being methods development.

I would, I'd stretch that out a little bit.

I'd say I saw that maybe being 30, 40% being methods development.

And then there was a.

Another 30% in the middle where it was pushing the boundaries of a method a little bit and then using it towards a biological question, even if they weren't completing the biological question.

Yeah.

I think the trend of applying to technologies in the fields of precision medicine or in general translational proteomics, as we call that, I think that evolution has started many years ago.

I just thought, which is much stronger, much more visible, this hoopoe than I had seen before.

Yeah.

I also think this hupo made a very strong effort to reach out beyond the human population.

We saw this in our chat with Jennifer Keddes McAllister where she shared about this one health concept where you're looking at agriculture, you're looking at crop science.

And that was very evident for me at World Hoopoe this year, that there was a reach much further than just human.

Yeah, in fact, I almost forgot about that.

And in fact, I very much appreciated those sessions that went beyond the human proteome, but looked into adjacent fields in the agriculture field.

In fact, the presentation, one of the plenary presentations on honeybees was actually really eye opening in that context.

Absolutely.

And I think that concept that we as humans are part of larger ecosystems and that the anti fungal agents that we give to our crops have an impact on our system and our ability to defend against funguses, for instance.

And so that that human, in the middle of a much larger set of things was, was really exciting to see for me and exciting to learn about.

So did you see any new trends or new developments that you want to share here?

Probably the largest surprise for me was hearing the word proteoform as much as I did.

It felt to me and I recognize that I'm knee deep in thinking about proteoforms from the work we're doing at Nautilus.

But despite that, I heard more people talking about proteoforms as an important thing to look at, as an important thing to measure as a driver of protein interactions.

I think there were some papers earlier this year where we groups looked at different splice variants and different PTMs and those actually completely changing the set of proteins that were interacted with.

And so for me my major take home was that this transition to awareness and interest in proteoforms really took off this year.

I hadn't seen, there was always interest work from people doing top down or native Ms. And but I think this year I saw just so much more of it everywhere.

Yeah, maybe I would phrase it differently.

I think, I think I saw the terminology of protoform or proteform groups escaping the top down field and being discussed in more general terms.

Right.

What does it mean to detect a protein?

Well, maybe we should be more specific about the real functional form that we are interested in and obviously that's the proteform.

So it was still a very small, you know, small number of presentations there.

But I think what I agree with you that those were highly attended and people were very interested in that as, you know, the next frontier in proteomics.

I think what's, what's missing in the conversation of the proteform is the connection to the function and I think that's what people are, you know, interested in.

And of course that takes more than just a measurement.

It usually takes a bigger study, it takes more context and I think that's what's going to be, you know, evolving over the next, you know, many, many years that people have the ability to measure more of the protein farms they're interested in and then get that functional context.

Absolutely.

I think, I think there's an awareness and Neil has been promoting for a long time that we need to have an atlas of proteform so we even know what's out there right now.

We don't even really know what the universe looks like.

We just think it is impossibly vast.

And so taking that one step further and saying, all right, not only asking the question what's there?

But how does it functionally contribute?

Does it have a different, does this particular form with these combinations of modifications have a different enzymatic activity or it binds to proteins at a different rate?

Those functional down downstream questions have been looked at by cell signaling people and biophysicists for a long time.

But bringing scale to that question, I think I just, I saw, I saw a lot more awareness of getting beyond the peptide.

One other thing I saw that was a little, I think has potential to be very confusing to folks outside of the field of proteomics, but was there, it is very clear what the definition of a proteoform is.

It's clear what the definition of a proteform group is in terms of proteform being defined by a specific chemical entity that has a combination of splice variants, truncations and multiple post translational modifications.

In order to measure a proteoform you really need to look at the full protein there.

But what we also saw was we saw some people who are looking at one PTM site for instance, and both in targeted immunoassays as well as in mass spectrometry and then saying, oh, we're measuring proteoforms.

No, you're measuring post translational modifications.

That's important, that's valuable.

But that is not the same as measuring the full proteoform or the proteform group.

And just from a nomenclature standpoint, right now that's already a little confusing within the field.

And I can only imagine to biologists not in the field that if we're not careful with our language that, that, that could, could really, really confuse people.

Certainly some more education to do in that respect, people using or utilizing the terminology.

But I think that will come as more and more people are focusing on, on this way of approaching proteins.

Right?

Absolutely.

And just recognizing what the, what the advantages and limitations are of each of the different technologies in terms of their ability to measure things like CO occurring PTMs.

That simply isn't something that can typically be done with a standard immunoassay or even a digested protein.

Where we're looking again at peptides, we, we may every now and again get lucky and have two phosphocytes on the same tryptic peptide, but that's certainly not the norm.

And if we want to look at multiple PTMs that are very far away from each other, they will not show up on the same peptidic fragment.

So just as a field, there's a little bit of as this is growing in awareness, this is growing in conversation and parlance, how can we make sure that, that we're each talking about the same thing?

Agree.

What else did you notice?

Well, a second big theme was a continuation of what we discussed, you know, earlier this year, but also last year, which is the, you know, the terminology of, you know, multi omics, the approach to take several technologies or several measuring modalities and essentially contextualize the results that you have obtained with proteomics so you understand more about the results that you have created in terms of biological functions.

And so that's trend, I think that's a multi year trend that continues.

And in that context, I think there's two things that are responsible for that.

One is certainly that from a vendor perspective, the vendors now provide instrumentation and methods that are much more integrated, but they're also standardized to the point where you don't spend a lot of time optimizing, for example, your LC gradient, you simply walk up and tell the instrument you want to do 20 runs per day and then everything is optimized for you and you move on.

Or for affinity based methods, you just say, here's the targets I'm interested in.

I need to run 500 samples.

The vendors have done their part of providing a more standardized readout so customers don't spend a lot of time optimizing methods for it to get an optimized method that enables now researchers to do more experiments utilizing different technologies.

I think that's part of the multi year trend we're going to observe where you now have the opportunity, a whole set of tools to actually dive deep into biology.

That's interesting.

So clearly I am a multiomics researcher and have been for a very long time, so I'm excited to see this, this trend grow.

Do you think that this is coming from?

So what you alluded to as the rationale for why this is growing is that the toolkit of proteomics is becoming more accessible.

To what extent do you think this is coming from the other side, where we have folks who are traditionally genomics researchers feeling bottlenecked or like there's a barrier in their ability to convert what their findings are to biological consequence and that there's a need for the Protein layer to be part of those full functional models.

That's an interesting question.

I think what I sense is that the initial vision of proteomics, which was very similar to genomics, right?

You've got to be able to just do the whole proteome.

You got to do it in half the time and one third of the cost.

So this, this type of technology mindset has basically, you know, disappeared.

It's now much more about I need to be able to understand functional biology.

And that means I need more than one tool because, you know, any tool set is limited given the complexity of the proteome.

And I think that's what drives it, that everybod understands.

I can only observe a sub portion of my science.

If I use one tool, I need to use several tools.

And I think that's what drives the multi omics.

And I think multi omics maybe gets a little bit overextended.

It's not just multiple modalities, but it's just multiple methods that often uses the terminology multi omics.

So in that context, maybe multi omics is broader than initially discussed.

Yeah, I guess my question was more about do you see that the push into multiomics is proteomics people moving closer to genomics and incorporating genomics in their studies or is it more genomics people moving closer to proteomics and incorporating proteomics in their studies?

I think I see both.

I mean, some vendors or some platforms utilizing more, you know, proteomic nucleic acid amplification type approaches.

So you might get people from the genomics field familiar with the technology of PCR or next generation sequencing, saying, hey, I can extend now my measurement, my tech tool set that I already know very well into the proteomics space.

And then they basically have access through these methods into the proteomics.

And I think that's happening in parallel.

While the proteomics tools are solid enough, high resolution enough, fast enough to create many observations that essentially cannot be really well understood until, unless you contextualize it with other modalities.

Interesting, interesting.

Yeah, I think one of my takeaways I think is related.

I guess I was aiming to have three takeaways for the year.

I ended up with four.

But two of them are very close to each other.

So one of my big takeaways was really experiment size that and, and scale of experimentation.

This year I saw a larger number of large scale, thousands of sample studies, tens of thousands of sample studies in proteomics.

There was the, the work that Sarah Hadi published earlier in the year where she did a multi, a multi platform comparison across thousands of samples.

We had UK biobank studies, we had another study that was, was promoted again on the scale of that is getting off the ground, aiming to measure potentially a million samples by proteomics.

And so that just that concept of experimentation on that, on that order, I saw so much more this year than I had in previous years.

But the, the flip side of that was I can't tell whether it is a small number of people who have done these massive, massive studies and whether that's the whole field is shifting towards things needing to be on that scale or if it's just it has become more possible for those, those, those couple, couple people to do things on that scale.

Interesting observation.

I think it aligns a little bit what I, you know, sort of mentioned with the multiomics, that it basically is a multiple platform discipline, proteomics now.

And I agree with you, I think there's not a lot of laboratories that actually are capable of very large studies because it entails more than just access to the platform.

It entails experimental design, the biostatistics, in the end access to a particular sample site, whether it's biobank sample or other.

So it's probably the fact that there's more focus around these type of projects enabled by technology.

I'm not sure this is yet mainstream that every laboratory is planning to do these type of studies.

Yeah, yeah, absolutely.

So I think for me this takeaway in scale and this growing desire for larger scale proteomics was something I observed.

I think that goes hand in hand with something that is of course top of mind in the world, which is AI.

And how as we think about proteomics, there's now a HUPO AI readiness group that is working on developing standards for, for proteomics to interact with AI tooling.

There have been some folks building foundation models in proteomics to complement foundation models in genomics.

And so I think that rise of studies incorporating late stage proteomics data into their, into their AI models is growing a lot.

And just to clarify what I mean by late stage proteomics data, what I'm really referring to is where in the integration process is the proteomics data brought in.

So there are different types of models where there's early integration, middle integration, late integration and early integration might be all the way starting from RNA SEQ data, raw reads being integrated at the spectra level to interpret the Actual spectra coming in and the actual primary data that comes out of that for subsequent analysis.

That's early stage integration where you're, you're integrating all the different data in the earliest part of the process.

Late stage integration, on the other hand, says we're going to take the simplest raw, the simplest, least raw, most processed outputs, protein abundances and quants, possibly after having been imputed, and merge those together with the again post processed RNA SEQ FPKM values that have already been cleaned up and normalized and potentially already looked at for which features are separated.

So it really, I haven't seen as much of proteomics data and other OHMS data being brought in together at the raw data stage.

It really has been at much later once you already have quants and maybe even already have differences between each of them.

And one of the potential benefits of, of moving earlier and doing early integration instead of late integration is that you might find things, for instance, there might be mutations that are found in the RNA SEQ data or the DNA SEQ data that can then go into the proteogenomic workflow to validate that those protein products are actually translated ones that contain those mutations.

You also might be able to do a better job of interpreting with the full data cube as opposed to just, hey, These are the 20 proteins that are different or the 20 transcripts that are different in the RNA data.

These are the 20 proteins that are different.

You could look at how the, how the entire collection across the whole system is changing from the transcript level to the protein level.

So there are sets of questions that are best answered by early integration.

But the data volumes are larger, the complexity of the analysis is substantially larger.

The domain knowledge that's required to do that is significantly larger.

So that's probably why most people don't do it.

But there are some major, major benefits to doing so.

The other place where AI was having an influence was inside so doing things like looking over spectral libraries to do a better job of identifying peptides, for instance.

So I think we are seeing AI creep into the ecosystem and not just be a consumer at the tail end of proteomics.

Yeah, maybe, you know, to the, to the topic of AI, I think I saw actually less posters on AI, but I think the observation that AI is now part of proteomics is absolutely correct.

In fact, I saw a poster from The Matthias Mann lab, where they essentially trained AI to learn all about the mistakes you can make, hoping that the next person who is going to use the platform is not making the mistakes and is in fact not wasting a sample because they know they're not ready for it.

And so I think that these are very interesting developments and certainly on the spectral side, I think, think we all can agree that machines are going to be much better reading spectra than humans will ever be, given the volumes that these instruments today generate.

I think just carrying that out, one of the, this segues into my last big observation from the year and then I'll ask you about your last observation in the year, which is really about quality.

And this is a place where I am both optimistic and concerned at the same time because these, as you said, there are a lot more people who are pressing the go button and generating DIA spectra and taking whatever the software tools come out and say, oh, this is how much of this protein is there based on this peptide roll up.

And those folks often do have a deep understanding that hey, I need to look at six daughter ions, I can't just look at one in order to be confident in a protein identification.

But then that identification leading to a protein group which may have multiple proteins associated with it, how that then gets boiled down to a single protein that says, hey, this is how much of this single protein is there.

There's, there's a chain, there's a tree, there's some complexity there that is, that is getting lost in the translation to a very simple table of this is the protein and this is how much is there that is used as the input for many other downstream analyses.

So there's, there's a little bit of a, an excitement that the field is having awareness about the need for improving quality and recognizing that not all protein quantification data is equal.

But on the other hand, I didn't see a clear set of solutions emerging to deal with the diversity of different types of proteomic workflows, proteomics data, whether it's TMT data or DIA MS.1 data or Spectral counting data or data from an affinity based platform, how we treat those all in common so that we can understand what are the quality aspects of the data.

Again, I saw a greater awareness of this as a field, but didn't see any major solutions coming forwards.

Yeah, I mean, I was also not seeing any solutions to the topic you addressed here.

I think there's always going to be as part of the method development, instrument development efforts, people who push the boundaries, but I am getting less and less concerned about it as we move forward.

Given that I see proteomics as a multi platform approach that if you're really interested in a target, you will get several ways to confirm your observations.

And so even if one platform gets it wrong or is not as confident as measurement, I think applying a second or third tool will ultimately flush out the truth.

It's just the part that's concerning is some of the experiments are very time consuming, whether it's a large study or it's an expensive experiment.

And so you might not want to spend a lot of energy on chasing a target that you created with a very, very expensive approach and then find later out that it is not a viable target.

And so in that sense I share a concern that we need to continue to make sure that we know what we're looking at.

Define what quality means, define what protein ID means.

And I think that goes back to just the conversation we just had around Proteform.

Right now, Proteform might be a terminology you easily throw around, but when it comes down to naming the proteform, we better be very clear and have a solid identification for the solid proof points for it.

Yeah, I saw a study earlier this year where in the study they were very excitedly commenting on that.

I think it was 60% of their quantifications had CVs below 30% and it was something on that order like, wow, all right, well what about the rest of the data and how should we think about it if the CVs are 100 or 200% and do we know that for most data sets and in many cases the answer is no.

So the fact that some protein IDs might be extremely confident, they might be extremely high quantitation accuracy, very high precision and other protein measurements in the same data set may not be.

I think that's something that is not well captured yet, but is an opportunity as a field for us to improve on.

So Andreas, what was your final takeaway from the year?

Well, final takeaway was that certain things that I didn't observe.

So for example, I was surprised that we didn't see more spatial proteomics.

That was a buzzword just two years ago.

And so yeah, there were some posters using very sophisticated multi imaging system now.

I mean they're really improving in resolution and certainly people make a lot of effort to look at biology relevant answers in these spatial efforts.

But I was surprised given all of the buzz around the technology that's been evolving the last 10 years and quite frankly the investment has been made in the field that we didn't see a lot more spatial proteom results there.

So it might be the fact that maybe the HUPO meeting is not the meeting that is focused typically on that.

But I would hope that in the future we'll see more of these technologies that are, you know, have spatial resolution that they actually also join, you know, the, the HUPO conference.

Yeah, absolutely.

I, I noticed the same thing.

I, I was expecting there to be this onslaught of spatial and single cell presentations and posters and I was surprised to see a lot less of that this year.

But clearly over the last several years it was a huge, huge emphasis and so maybe this is just a slight wiggle and next year it'll be all single cell and spatial all the time.

We'll find out.

But I also noticed a little bit less, less focus on that area this year.

Yeah.

So overall my takeaway for the, for, you know, the year but also for the conference was that preomic's landscape is obviously consolidating around clinical and translational translational applications while expanding the multi platform aspect of, and the new innovations around that.

And I think in general the field is maturing from an instrumentation driven discovery to workflow that are integrated and focused on biology.

A year later we're going to come back to this.

Well, I would hope to see true applications of AI where we can now say, look, those spectra are, weren't able to interpret now have solid evidence or we have a new way to impute missing data, for example.

So I hope that there is going to be some kind of breakthrough in this area.

I think the other area I'm really expecting to continue to evolve is the area we just discussed on proteoforms.

Looking much more at the functional aspect of the proteome.

I think the areas that I see continuously underrepresented structural biology, chemical chemoproteomics, protein, protein directions.

I think that these are, these are areas that I hope we will see many, many more breakthroughs in the future.

That's super.

Thank you.

Thank you very much.

Oh golly, I was hoping just to put you on the spot.

I'm, I'm going to go with, I do think this question of multi omics and how do we integrate multiple data types is going to continue to grow.

I think it's clear that there are, there's so much happening at different scales and the tooling to be able to ask those questions has gotten a lot better.

So I very much am expecting an expansion in multi omics studies.

I think we also see things like the CPTAC data sets are now just massive.

The number of cancers they've looked at, the number of different tissue types and the metadata around those samples is tremendous.

And so I'm anticipating that.

One of the other things we're going to see this coming year alongside this growth in AI is an increase in data reuse.

Now that we have these massive corpuses of data, some of these larger clinical studies, these tens of thousands of samples clinical studies.

And I think number three, I really do think this connection to biology and having more and more of these stories come out where it's so clear that it is that the proteome was a necessary part of the biological mechanism, I think, I think that's my other hope.

Thank you so much, Andreas.

That was a really fun conversation.

Always great to reflect and look back on the year.

We hope you enjoyed the Translating Proteomics podcast brought to you by Nautilus Biotechnology.

To contact us or for further information, please email Translating Proteomics at Nautilus Bio.