Customized Ligands: Making a 'Tailor-Made' Key for Cells

Imagine you walk into a clothing store.

The ready-made clothes come in sizes S, M, and L, but you have broad shoulders, a slim waist, and long arms—so something always feels off when you try them on. So you go to a tailor, get your shoulder width, waist, and arm length measured, and get a piece that fits you perfectly.

In the molecular world, the same logic applies.

Those molecules that can precisely find and bind to a target are called ligands—like a key that fits a specific lock. But the "ready-made keys" that nature provides often don't fit perfectly. So humans learned to "make our own keys"—this is what we call customized ligands.

1. Why customize? Are natural ligands not good enough?

Natural Ligands

Nature has spent billions of years crafting countless intricate ligand molecules. For example, insulin in your body can precisely find insulin receptors on the surface of cells, making the cells 'open the door' to absorb blood sugar. Or antibodies, which can recognize invading viruses and bacteria, marking them for clearance by the immune system.

These natural ligands are already amazing, but they all share one common 'design goal'—to survive and reproduce. All of their characteristics have evolved around this goal of survival and reproduction, which doesn’t necessarily align with human medical needs.

Here are a few of the most common ways they don’t quite fit:

Problem 1: Half-life is too short. Many cytokines have a half-life of only a few minutes to tens of minutes in the body. For example, IL-2, a cytokine that can activate immune cells, could theoretically be used to fight cancer, but it gets metabolized almost immediately in the body before it has time to work. It’s like a key that corrodes before you even turn it in the lock.

Problem 2: Not specific enough. Some ligands are very 'well-meaning'—they can bind to several similar receptors at once. This might be good in evolution, but a disaster for treating disease: you only want to activate receptor A for therapy, but it ends up activating receptor B as well, causing a bunch of side effects. It’s like a key that opens several locks at the same time.

Problem 3: Unstable conformation. Many ligand proteins are like Transformers, switching between different shapes. Viral fusion proteins are a classic example: sometimes in a 'pre-fusion' state, other times in a 'post-fusion' state. You want to make a vaccine that induces antibodies against the 'pre-fusion' state, but it keeps changing to the 'post-fusion' form, making your efforts useless. It’s like a key whose teeth keep changing shape—you align it with the lock but it’s never the right shape.

Problem 4: Size limitations. This is a practical constraint from the application scenario, not a flaw in the ligand itself. Antibodies are among the best ligands, but they are too big to cross the blood-brain barrier or penetrate deep into solid tumors. It’s like a big key—you want to carry it in your pocket, but it’s just too bulky.

So, it’s not that natural ligands are bad—they’re designed for nature, not for you.

2. What exactly does ‘customization’ change? Three levels of modification

The degree of customization for ligands can be divided into three levels, from shallow to deep. It's like tailoring clothes—sometimes you just adjust the pant length, sometimes you change the fabric, and sometimes you make an entirely new outfit from head to toe.

Level 1: Change details—to make it 'fit better'.

The most basic customization involves point optimization on the sequence of a natural ligand. It's like having a generally fitting piece of clothing but the waist is a bit loose or the sleeves are a bit long—you just tweak a few spots. Ligands work the same way: some residues at the binding interface with the target are especially critical (these are called 'hotspot residues' and contribute most of the binding energy, usually can’t be mutated freely). By optimizing the nearby minor residues to fill gaps or add extra interactions, you can increase affinity by one to two orders of magnitude.

Even more important is selective modification. If a key can open several locks, you just file down the less critical parts so it only opens the lock you want. In ligand engineering, this is called 'specificity modification'—going from 'broad binding' to 'precise recognition'. This first level of modification is the most mature, with antibody drug affinity maturation being a typical example.

Level 2: Change structure—to make it 'stay in shape'.

If the first level is 'patching and sewing', the second level involves changing the structure. Often the biggest problem with ligands isn’t 'not binding tightly enough' but 'unstable shape'—sometimes it's like this, sometimes like that. So what do you do? Pin it down.

The most common 'pins' are two types: disulfide bonds, where you introduce cysteines at two nearby points in the natural protein conformation to form a covalent disulfide bond, like rivets holding the structure in place; and proline mutations, because proline is the 'stiffest' amino acid, so replacing flexible residues with proline in a certain region can 'lock in' that part. This is called conformation locking—fixing the ligand in the functional state you want.

Besides internal structural fixes, this level also includes optimizing the molecular assembly form—multivalent display. One key opens one lock, but what if you line several keys together? They can insert into a row of locks simultaneously, multiplying the binding force. This is why vaccines often display antigens on nanoparticles—the multivalent effect can boost immunogenicity several to dozens of times, significantly increasing antibody titers.

Level Three: Built from Scratch — Completely "Tailor-Made."

This is the most groundbreaking and cutting-edge level: designing from scratch. Instead of starting from any natural protein, computational methods are used to "draw" an entirely new protein molecule from zero, making it just right to bind to the target you want.

You might ask: how is this even possible? Simply put, scientists first figure out what the "binding surface" on the target protein looks like, then use computational design software to "stack" a protein scaffold on this surface so that its shape, charge, and hydrophobicity perfectly complement the target — like scanning your foot and 3D printing a pair of shoes that fit perfectly.

How amazing are these de novo designed "mini ligands"? Tiny — only a fraction (one-tenth to one-hundredth) the size of an antibody; stable — some de novo designed mini proteins can recover their activity even after being heated to 95°C; strong — some top-performing mini ligands can reach nanomolar or even picomolar affinity.

3. So, what’s the point of custom ligands?

Applications of Customized Ligands

After saying all this, what exactly can these 'custom keys' be used for? Actually, you might have already come across them.

**Biopharmaceuticals: More precise drugs**

- **Antibody drugs:** Almost all marketed antibody drugs have undergone customization—humanization, affinity maturation, Fc engineering… PD-1 antibodies and HER2 antibodies you've heard of are all customized ligands.

- **Cytokine engineering:** For example, long-acting interferons (extended half-life through PEGylation or Fc fusion), biased IL-2 (activates only anti-cancer immune cells without activating inhibitory ones).

- **Bispecific antibodies:** Artificially created 'double-headed keys,' one end binds to cancer cells, the other to immune cells, bringing immune cells to cancer cells to mediate killing.

**Diagnostic testing: More sensitive 'probes'**

The antibody in COVID-19 antigen test kits that recognizes the virus is a customized ligand. The higher its affinity and specificity, the higher the detection sensitivity—even very low virus concentrations can be detected.

**Synthetic biology: More controllable 'switches'**

In synthetic biology, customized ligands can be used to design various 'molecular switches'—for example, activating an enzyme only in the presence of a specific small molecule, or triggering gene expression only when cells detect a specific signal.

4. How AI Accelerates 'Customization': The Role of MatwingsVenus™

After explaining all the benefits of customized ligands, you might ask: Is it difficult to make a 'molecular key'?

The answer: The technical barrier has always been high, and the difficulty persists. In traditional R&D, not only is it hard to develop, but the cycle is also extremely long. Adding AI technology can greatly reduce trial-and-error costs and significantly speed up the entire customization process.

Traditional ligand customization relies heavily on trial and error: early-stage random mutagenesis requires building tens of thousands to hundreds of thousands of mutant libraries for screening; later-stage structure-based semi-rational design narrows the range, but still needs a lot of experimental verification—the process often takes years and costs a fortune.

AI is changing the game. AI protein design platforms like MatwingsVenus™ (Xiaowu™) from Matwings Technology are shifting ligand customization from 'high-throughput trial and error' to 'rational inference.'

4.1. First, see exactly what the “lock looks like” — structure prediction and interface analysis. To make a key, you need to first see what the keyhole looks like. MatwingsVenus™ (Xiaowu™) can predict the 3D structure of a target protein from its amino acid sequence and can also predict how the ligand-target complex binds, marking the key residues at the binding interface. This means you know which positions are the most important, so you only need to focus mutations on these sites — shrinking the screening space from the “entire interface” to just “a few key residues,” and you can imagine how much more efficient that is.

4.2. First, figure out “which modifications are reasonable” — pre-screening the effects of mutations. The traditional approach is: synthesize dozens of mutants and test them one by one in the lab. Most mutants either don’t help or actually make things worse. MatwingsVenus™ (Xiaowu™) can evaluate the effect of each mutation computationally — how much would it improve affinity? Will it disrupt protein folding stability? Will it also increase binding to homologous off-targets? Basically, before you make any mutant, the AI runs a “virtual screen” to rule out the obviously bad ones and pick out the candidates with the highest chance of success for you.

4.3. Stabilize the backbone first, then optimize function — the “stabilize first, optimize later” roadmap. Affinity, specificity, and stability form a triangle balance. Often, scientists instinctively try to boost affinity first and then stabilize, but experience shows that “stabilize the backbone first, then optimize function” is usually more efficient. MatwingsVenus™ (Xiaowu™) can systematically assess which mutated sites can enhance overall stability, which sites are suitable for functional optimization, and which mutation combinations are compatible — basically giving you a modification roadmap before you even start experiments.

4.4. Structural reference for conformational locking. For dynamic ligands, MatwingsVenus™ (Xiaowu™) can predict the 3D structure of the ligand in its main functional conformations, providing reference points for identifying key structural constraint sites. Researchers can then design disulfide bonds, proline mutations, or other conformation-locking strategies. It’s like securing a key that easily deforms — the AI first marks the best spots to add stabilizing structures, but the researcher decides how to implement it.

With the “dry-wet closed-loop” R&D system built on MatwingsVenus™ (Xiaowu™), meaning tight integration of computational design and experimental validation, protein development cycles can be reduced from the traditional years to just months, experimental trial-and-error is greatly reduced, and R&D success rates are significantly improved. It’s not that AI replaces experiments — it just narrows the trial-and-error range by an order of magnitude, letting limited experimental resources hit where they matter most.

5. The “impossible triangle” of custom ligands

Triangular Balance of Customized Ligands

After talking so much about the benefits, is a customized ligand better the more perfect it is?

It's not that simple. In ligand engineering, there's an age-old challenge—the triangle balance of affinity, specificity, and stability. It's really hard to push all three to the extreme at the same time. For example: if you want stronger binding, you need more hydrophobic interactions, but too many hydrophobics can make the protein itself prone to aggregation, lowering stability; if you want a larger, more complementary binding surface, you might also end up binding other homologous proteins, reducing specificity; if you want the protein to be particularly stable, you make the backbone really 'rigid', but if the ligand needs the target to undergo an induced-fit conformational change, too much rigidity can hinder the adjustment, actually lowering affinity.

So experienced designers don’t chase 'perfection,' but look for a balance point—deciding what to prioritize and what to compromise on based on the specific application.

It’s a bit like making clothes—you want it super slim-fit, and it might not be comfortable to move in; you want it super comfy and durable, and the fit might not be that precise. There’s no absolute 'good,' only what fits your needs.

The role of AI isn’t to break this triangle balance for you, but to help you find that optimal balance faster and more accurately.

Conclusion

From discovering natural ligands, to modifying them, to designing completely new ligands from scratch—the mastery humans have over molecular recognition is deepening step by step.

Before, we could only pick ready-made keys from nature’s 'shelf,' and being able to fit a lock was already good enough; now, we can take the target’s 'lock template,' draw our own blueprint, polish it ourselves, and create a perfectly fitting key; and AI makes this customization cycle shrink from years to months—from 'the master slowly polishes' to 'AI makes a precise prototype.'

This is the story of customized ligands—molecular recognition is no longer just a gift from nature, it can also be achieved through human rational design; and AI makes this design faster, more accurate, and more accessible.

One day, when we can quickly 'customize' a perfect key for any protein or site, that will be the true 'kingdom of freedom' in biomedicine and synthetic biology.