Back in the old, grimy chapters of drug discovery, most medicines worked like bouncers - block a protein, shut down a pathway, call it a night. For decades, researchers kept running into the same dead end: huge chunks of biology refused to cooperate because many disease-driving proteins had no neat little pocket for a classic drug to jam itself into. Attempt after attempt failed on these so-called "undruggable" targets. And then the case took a strange turn. What if, instead of blocking a protein, you convinced it to meet someone?

The suspicious rise of the molecular matchmaker

That is the central mystery in Fan et al.'s review on proximity-inducing drugs - molecules designed not just to inhibit proteins, but to bring proteins together or stabilize interactions between them Fan et al., 2025. It sounds almost absurdly social for biochemistry. Proteins, as it turns out, can be manipulated like guests at a chaotic wedding reception: seat the right two people together and suddenly the whole evening changes.

You may have heard of PROTACs, the celebrity branch of this family. They recruit a cellular cleanup crew, often an E3 ubiquitin ligase, to tag a target protein for destruction. But that is only one act in a much weirder show. Proximity-inducing drugs can also alter signal transduction, gene transcription, chromatin regulation, and protein trafficking. Translation: instead of merely turning one switch off, they can rewire entire conversation networks inside the cell.

That matters because biology is less like a row of on-off buttons and more like a group chat where nobody stops replying.

The crime scene: too many proteins, not enough clues

Here is the problem. If you build a molecule that forces proteins into awkward proximity, you need to know three things:

1. Who is it actually binding?
2. What downstream chaos does it cause?
3. Can you find suitable ligands to build these molecular matchmakers in the first place?

This is where proteomics enters the story, flashlight in hand.

Proteomics is the large-scale study of proteins - what is present, how much of it is around, how it changes, and who is interacting with whom. If genomics gives you the cast list, proteomics shows you who is sneaking around backstage at 2:47 AM with motive and opportunity. Modern mass spectrometry-based proteomics lets researchers map drug targets, identify off-target effects, and track system-wide consequences after treatment.

Fan and colleagues argue that proteomics now sits at the center of rational proximity-drug discovery because it can reveal both the direct target engagement and the proteome-wide downstream response Fan et al., 2025.

Follow the protein fingerprints

A few recent reviews and studies back this up. Békés et al. described the broader promise of targeted protein degradation and proximity pharmacology, showing how these methods expand what counts as druggable biology (Nature Reviews Drug Discovery). Schreiber also framed "proximity pharmacology" as a wider strategy beyond degradation - one that includes induced interactions for regulation, not just destruction (Cell).

Meanwhile, chemoproteomics and ligandability mapping have become key tools for finding proteins that can actually be recruited into these schemes. Recent work on proteome-wide ligand discovery has pushed researchers closer to a map of which proteins carry chemically tractable footholds, a bit like discovering half the houses in town secretly have side doors nobody documented properly.

And yes, machine learning has started showing up to the investigation with one giant evidence board.

The algorithm in the trench coat

Fan et al. highlight a growing role for machine learning on proteomic data - especially for predicting targetable proteins, interpreting massive protein-response datasets, and identifying candidate ligands Fan et al., 2025. This makes sense. Proteomics generates messy, high-dimensional data at a scale that can make ordinary analysis cry softly into a spreadsheet.

ML models can help detect hidden patterns in protein abundance changes, interaction networks, and ligandability signals. They are not magic, obviously. They are more like overcaffeinated interns with very good pattern-recognition skills and occasional delusions of certainty. Still, paired with high-quality datasets, they can narrow the search space dramatically.

If you're trying to visualize these tangled interaction networks, tools built for structured thinking - like mapb2.io - are the sort of thing that can help you sketch the web without making your notes look like a conspiracy wall from a detective drama. Which, to be fair, this field increasingly resembles.

Why this gets interesting fast

The real excitement here is not "we found another inhibitor." It is that proximity-inducing drugs may let researchers expand the target space into proteins and pathways that traditional small molecules handle poorly. Instead of asking, "Can I block this protein?" the better question becomes, "Can I make this protein do something new by changing who it interacts with?"

That could matter in cancer, neurodegeneration, immune disorders, and other diseases where cellular behavior depends on dynamic protein networks rather than one obvious bad actor.

But let's not start writing victory music yet. The field still faces hard problems: limited ligand availability, uncertain selectivity, incomplete understanding of induced protein complexes, and the usual reproducibility headaches that stalk modern biomedical research like a villain who simply refuses to stay dead.

The verdict, for now

This review does not present one flashy silver bullet. It presents something better: a map of an emerging strategy. The clue trail points toward a future where proteomics plus machine learning plus proximity pharmacology gives drug discovery a new playbook - one built around orchestrating protein relationships, not just blocking them.

That is a subtle shift, but a big one. Drug design used to ask proteins to stop talking. Now it may start arranging introductions.

And honestly, in a field where half the struggle is getting molecules to behave, playing molecular matchmaker might be the weirdly brilliant move nobody should have bet against.

References

1. Fan R, Ni J, Zhou T, Jiang H, Zhao W, Tan M. Proteomics-Driven Strategies for Proximity-Inducing Drug Discovery. Angew Chem Int Ed Engl. 2025. DOI: 10.1002/anie.3307512. PubMed: PMID 42294530

2. Békés M, Langley DR, Crews CM. PROTAC targeted protein degraders: the past is prologue. Nat Rev Drug Discov. 2022;21(3):181-200. DOI: 10.1038/s41573-021-00371-6

3. Schreiber SL. The rise of molecular glues. Cell. 2021;184(1):3-9. DOI: 10.1016/j.cell.2020.12.020

4. Bondeson DP, Crews CM. Targeted protein degradation by small molecules. Annu Rev Pharmacol Toxicol. 2023;63:347-370. DOI: 10.1146/annurev-pharmtox-042122-024553

5. Spradlin JN, Hu X, Ward CC, et al. Harnessing the anti-cancer natural product nimbolide for targeted protein degradation. Nat Chem Biol. 2019;15(7):747-755. DOI: 10.1038/s41589-019-0304-8

Disclaimer: This blog post is a simplified summary of published research for educational purposes. The accompanying illustration is artistic and does not depict actual model architectures, data, or experimental results. Always refer to the original paper for technical details.