Affinity Selection Mass Spectrometry: Advances in Drug Discovery

Affinity selection mass spectrometry (AS-MS) has become one of the more powerful tools in early-stage drug discovery because it lets researchers fish for drug-like molecules that bind to a biological target without needing fluorescent tags, radioactive labels, or any other artificial modifications. The technique works by mixing a target protein (or even an RNA) with thousands of candidate compounds at once, separating the molecules that stick from the ones that don’t, and then using a mass spectrometer to identify the binders by their molecular weight. Over the past decade, advances in separation methods, automation, and target scope have pushed AS-MS well beyond its original niche, making it relevant to fragment screening, membrane proteins, RNA targets, natural products, and even covalent drug design.

How the Screening Works

The basic idea behind AS-MS is straightforward: combine a biological target with a pool of small molecules, give them time to interact, then physically separate the bound complexes from everything else and use mass spectrometry to figure out what stuck. The differences between AS-MS methods come down to how that separation step happens. Three approaches have emerged as the most widely used. Pulsed ultrafiltration (PUF) passes the mixture through a membrane that retains large protein-ligand complexes while letting unbound small molecules wash through. Size-exclusion chromatography (SEC) separates by molecular size on a column. Magnetic microbead affinity selection screening (MagMASS) takes a different route by anchoring the target protein onto magnetic beads, then pulling the beads (with bound ligands attached) out of solution using a magnet.¹

PUF and SEC are both solution-phase methods, meaning the target protein floats freely in buffer during the incubation, which closely mimics physiological conditions. MagMASS trades some of that naturalness for practical convenience: immobilizing the target on beads simplifies the wash steps and makes automation easier. Among these, the size-based separation approaches have shown the broadest success across different types of drug targets, largely because they mesh well with liquid chromatography-mass spectrometry (LC-MS) workflows that pharmaceutical labs already have in place.¹

Why Labels Are Not Needed

Most high-throughput screening methods in the pharmaceutical industry rely on some kind of reporter signal to tell you whether a compound is active. That usually means attaching a fluorescent or radioactive tag to the target, a competitor ligand, or both. Those labels can interfere with the binding event you’re trying to measure, and some target classes are genuinely difficult to label without disrupting their function. AS-MS sidesteps the problem entirely. Because the mass spectrometer identifies compounds by their inherent molecular weight, there is no need for any chemical modification of the target or the library compounds. The technique works with all classes of receptors, enzymes, and ligands, and it is compatible with a wide range of buffer conditions and cofactors.²

This label-free nature has a practical knock-on effect: you don’t need to develop a custom assay for every new target. Traditional high-throughput screening campaigns often spend months validating a reporter assay before the first compound is ever tested. With AS-MS, the same general protocol works across targets, which compresses timelines considerably. That versatility is a big part of why adoption has grown in industry settings where speed matters.

Screening Fragment Libraries

Fragment-based drug discovery starts with very small molecules, typically under 300 daltons, that bind weakly to a target. The idea is to find tiny chemical footholds and then build them up into larger, more potent drug candidates through medicinal chemistry. The challenge is that weak binders are hard to detect with many standard assay methods, and the libraries involved can be large. Techniques like surface plasmon resonance and nuclear magnetic resonance (NMR) are commonly used for fragment screening, but they tend to consume substantial amounts of purified protein and are relatively slow per compound tested.

AS-MS has proven well suited to fragment screening because it can process compounds in large pools and uses far less protein per experiment. In one study targeting a G protein-coupled receptor (GPCR), researchers used an affinity MS-based fragment screen that reduced protein and compound consumption by two to four orders of magnitude compared to standard methods, while also speeding up data collection by two to three orders of magnitude. That campaign identified a new negative allosteric modulator of the adenosine A2A receptor bearing a chemical group (an azetidine) that had not been seen before in that target’s pharmacology. The compound was predicted to occupy a sodium-binding pocket on the receptor, a challenging allosteric site that conventional screening methods would be unlikely to probe efficiently.³

The ability to screen fragments rapidly and cheaply matters because the chemical diversity within a fragment library is enormous relative to each molecule’s simplicity. Picking up weak but genuine binders from a sea of inert molecules is fundamentally a signal-to-noise problem, and AS-MS handles it by relying on affinity-based enrichment rather than measuring a proxy signal like fluorescence intensity.⁴

Membrane Proteins and GPCRs

Membrane proteins, especially GPCRs, are among the most important drug target classes in medicine but are notoriously difficult to work with in screening assays. They sit embedded in cell membranes, and pulling them out into solution often disrupts their structure and function. Many screening platforms require purified, soluble protein, which means researchers sometimes use thermostabilized, mutated versions of a receptor that are easier to handle in the lab. That creates a subtle problem: the stabilized receptor may be locked into one conformation and unable to engage compounds that would bind to a different, biologically relevant state.

AS-MS offers a workaround because some implementations can screen against targets in their native membrane environment. In work on the glucagon-like peptide-1 receptor (GLP-1R), a class B GPCR involved in blood sugar regulation, researchers ran an affinity MS screen using cell membranes containing the wild-type receptor and identified 18 confirmed binders. When they repeated the experiment using membranes expressing a thermostabilized, crystallography-grade mutant of the same receptor, every single one of those ligands lost its binding. The stabilized receptor was locked in an inactive conformation and simply could not engage the compounds that interacted with the active form.⁵

That finding underscores a broader point about drug screening: the more artificial the experimental setup, the more likely you are to miss real hits or find ones that don’t translate to biology. Screening against wild-type membrane-embedded receptors, while technically harder, captures a more complete picture of which molecules actually interact with the target in a physiologically meaningful way.

Targeting RNA with Small Molecules

For decades, drug discovery focused almost exclusively on proteins. RNA was considered too floppy and structurally unpredictable to be a viable small-molecule target. That view has shifted as researchers have found structured RNA elements involved in disease, from bacterial riboswitches to oncogenic noncoding RNAs. But the toolbox for screening RNA targets has lagged behind what exists for proteins. AS-MS has recently been adapted to fill that gap.

Using a platform called ALIS (Automated Ligand Identification System), researchers screened 42 different RNA targets from various classes against roughly 50,000 chemically diverse drug-like compounds and an additional 5,100 tool compounds with known biological activities. Of the 42 RNAs tested, 40 bound at least one small molecule. On average, each RNA target bound only about 0.04% of the tool-compound library and 0.01% of the diversity library, confirming that the hits were selective rather than promiscuous stickers.⁶ The very low hit rates are actually encouraging: they suggest that the compounds binding each RNA are doing so through specific structural recognition rather than nonspecific electrostatic attraction, which has historically been the concern with RNA-targeted screening.

This approach has also been used to discover bioactive compounds against noncoding RNAs. In one campaign, AS-MS identified small molecules active against Xist RNA, a long noncoding RNA involved in X-chromosome inactivation.⁷ That the ALIS technique, originally developed for protein targets, could be adapted to RNA without a complete overhaul of the workflow speaks to the flexibility of AS-MS as a platform.⁸

Prospecting in Natural Products

Natural product extracts, such as plant or microbial extracts, are some of the richest sources of chemical diversity on the planet, but they present a nightmare for conventional screening. A single crude extract can contain hundreds or thousands of individual compounds, many at very low concentrations, and the active ingredient may be a tiny fraction of the total mixture. Traditional bioassay-guided fractionation works but is slow: you fractionate the extract, test every fraction for activity, fractionate the active fraction again, test again, and repeat until you isolate the active compound. That process can take months.

AS-MS shortcuts this by letting the biological target do the sorting. You incubate the crude extract with the immobilized target, wash away everything that doesn’t bind, and identify the retained molecules by mass spectrometry. In a recent study targeting an enzyme from the parasite that causes schistosomiasis (Schistosoma mansoni), researchers used magnetic nanoparticles coated with the target enzyme and screened a crude natural product extract. The AS-MS step flagged six putative ligands with affinity ratios between about 1.5 and 2.9. Analytical microfractionation then pinpointed the main bioactive region to a narrow chromatographic window overlapping with two of those ligands, one of which was tentatively identified as asperphenamate.⁹

What makes this approach powerful is that it collapses what used to be an iterative months-long fractionation campaign into a single affinity-enrichment step followed by targeted analytical work on only the regions that matter. The rest of the extract, which may contain interesting chemistry but not relevant binders, is set aside immediately rather than occupying lab time.

Covalent Binders and Molecular Glues

Most AS-MS work assumes that compounds bind to their target reversibly, forming non-covalent complexes that can be separated and detected. But some of the most exciting drug classes in recent years form permanent or semi-permanent covalent bonds with their targets. Covalent inhibitors like osimertinib and sotorasib have transformed cancer treatment, and molecular glues, which work by stabilizing interactions between two proteins that wouldn’t normally form a complex, represent a fast-growing area of pharmacology.

Screening covalent binders with AS-MS introduces a technical challenge: if the compound is permanently stuck to the protein, the standard separation and detection workflow may not cleanly reveal the bound ligand’s identity. A recent advance addressed this by developing a reversibly reactive affinity selection platform called ReAct-ASMS, specifically designed for covalent peptide binders. The system uses tandem high-resolution mass spectrometry to identify covalent binders enriched from synthetic peptide libraries against native protein targets, providing a label-free and tunable way to discover covalent pharmacology that did not previously exist in AS-MS form.¹⁰

On the molecular glue side, native mass spectrometry has been combined with collision-induced dissociation to study how glue-mediated ternary complexes (target protein + glue + partner protein) form and fall apart. Researchers examined molecular glue-induced complexes between mTOR-related proteins and FKBP12, as well as a PROTAC-induced complex between FKBP51 and an E3 ligase called VHL. The native MS approach allowed them to measure not just whether the complexes formed but also the pathways by which they disassembled, information that is valuable for optimizing glue potency and selectivity.¹¹

Native Mass Spectrometry and Quantitative Binding Analysis

Standard AS-MS workflows detect which compounds bind but don’t always tell you much about how tightly they bind or what’s happening at the molecular level. Native mass spectrometry, which uses gentle ionization conditions to preserve non-covalent complexes intact in the gas phase, adds a layer of structural and quantitative information. By keeping protein-ligand complexes intact during the measurement, native MS can directly observe all species in equilibrium within a solution: free protein, free ligand, and bound complex simultaneously.

An extension of this idea, collision-induced affinity selection MS (CIAS-MS), incorporates controlled energy to break apart complexes in a stepwise fashion. The quadrupole mass filter first isolates the complex of interest, then a defined collision energy dissociates it so the released ligand can be detected and quantified. This approach gives researchers a way to rank binders by relative affinity and to study binding in complex mixtures where multiple ligands may compete for the same site.¹²

Practical Challenges and Non-Specific Binding

AS-MS is not without its headaches. The most persistent one is non-specific binding: compounds that stick to the protein surface, the membrane, or the beads through generic hydrophobic or electrostatic interactions rather than fitting into a functional binding site. Every AS-MS experiment produces some of these false positives, and distinguishing them from genuine hits is one of the trickiest parts of the workflow.

The choice of membrane in ultrafiltration-based AS-MS is critical. The membrane needs to retain the large protein-ligand complexes while letting unbound small molecules pass through cleanly. If the membrane itself is sticky, you get background signal that muddies the results. Control experiments run without any protein present are standard practice for flagging nonspecifically binding compounds. Replacing membranes between assays can also reduce cross-contamination, especially when significant background binding is observed.¹³

Beyond membrane effects, compound compression is another issue. When thousands of compounds are pooled together for screening, some share similar molecular weights. If two compounds have nearly identical masses, the mass spectrometer may not be able to resolve them, leading to ambiguous hit calls. Recurring background signals from plasticizers, detergents, or other lab contaminants can further complicate data interpretation. These are solvable problems but require careful attention to data processing, and teams running large-scale AS-MS campaigns typically invest significantly in custom informatics pipelines to handle hit filtering and false-positive mitigation.

How AS-MS Complements DNA-Encoded Libraries

DNA-encoded library (DEL) technology, where each compound in a combinatorial library carries a unique DNA barcode, has become another major platform for hit finding in pharma. DEL screens can interrogate billions of compounds in a single experiment, dwarfing the pool sizes practical for AS-MS. But the two approaches have different strengths. DEL screening detects binding through DNA sequencing enrichment and can miss compounds that bind weakly or transiently. It also requires the target to be immobilized on a solid support, which can introduce the same conformation-locking problems mentioned for GPCRs.

AS-MS, by contrast, works in solution and detects binders directly by molecular weight, providing orthogonal confirmation. Several groups have begun combining the two: using DEL to generate a broad initial hit list from a very large chemical space, then using AS-MS to validate and rank hits in a solution-phase, label-free setting. The synergy between the two technologies has been recognized as a way to combine DEL’s scale with AS-MS’s biophysical rigor.⁴ This pairing is especially valuable for targets where false positives from either method alone might derail a campaign.

Screening Scale and Compound Pooling

One of AS-MS’s practical advantages is its ability to screen compounds in pools rather than one at a time. A typical setup might mix a target protein with 1,000 to 2,500 compounds per well of a microplate, then analyze each well by LC-MS. Because the mass spectrometer distinguishes compounds by molecular weight, co-incubation does not create the interference problems that plague fluorescence-based assays when compounds absorb or emit at overlapping wavelengths.

The pooling strategy means that a library of 500,000 compounds might require only a few hundred to a few thousand wells, making the screen feasible on timelines of days to weeks rather than months. The compound pools do need some curation: groups of molecules with identical or near-identical masses should be placed in different pools so the spectrometer can tell them apart. But with modest informatics effort, pooling transforms AS-MS from a boutique technique into something that can operate at a scale competitive with more established screening platforms.¹⁴

The fact that binding detection relies on molecular weight also means that AS-MS works for chemotypes that might be invisible to other assay types. Compounds with no chromophore, no fluorescent handle, and no particular structural feature other than their mass are all fair game. This matters because medicinal chemistry often generates intermediates and analogs that lack the optical properties needed for activity-based assays but may still be pharmacologically interesting.