AbLecs and AI: Glycan Checkpoints in Cancer Care

A lot of immunotherapy strategy still assumes one core idea: if you release the brakes on T cells (think PD-1/PD-L1), the immune system will do the rest. That works—until it doesn’t. Many tumors don’t just hide. They actively signal immune cells to stand down.

A new modality highlighted this week (published December 16, 2025) puts a spotlight on a less-discussed set of “brakes”: glycan-mediated immune suppression, where sugar patterns on tumor and immune cells engage inhibitory receptors and dampen anti-tumor activity. The approach is called antibody–lectin chimeras (AbLecs), designed to bind and block these glycan signals. Early reports show AbLecs can outperform conventional antibody therapies in vitro and in vivo.

For teams working in the AI in Pharmaceuticals & Drug Discovery space, the real story isn’t just a new immunotherapy concept. It’s that glycan targeting is exactly the kind of high-dimensional, context-dependent design problem where AI-driven drug discovery can compress timelines—especially for engineering binding domains, predicting off-target risks, and optimizing developability.

Glyco-immune checkpoints: the suppression pathway many teams ignore

Answer first: Glyco-immune checkpoints are inhibitory interactions driven by glycans (complex sugars) that can suppress immune cell activation in tumors, similar in concept to PD-1/PD-L1 but mediated through carbohydrate recognition.

Cancer cells often present abnormal glycosylation patterns—extra sialylation, altered branching, unusual glycan motifs. These aren’t just decorative surface features. They can engage immune inhibitory receptors such as Siglecs (sialic acid–binding immunoglobulin-type lectins) on macrophages, NK cells, and other immune populations, pushing them toward tolerance instead of attack.

Here’s why this matters operationally for oncology programs:

• Checkpoint resistance isn’t always protein-protein. A tumor can look “PD-L1 low” and still be profoundly immunosuppressive.
• Innate immunity is a central bottleneck. Macrophages and NK cells often decide whether T cells ever get a real chance.
• Glycans change with context. Tumor type, stage, treatment pressure, and microenvironment all reshape glycosylation—making static targeting strategies brittle.

If you’ve ever looked at clinical data and thought, “Why did this patient with a decent T-cell signature still fail checkpoint blockade?” glyco-immune checkpoints are one of the more practical explanations.

What AbLecs are—and why the modality is interesting

Answer first: AbLecs are engineered proteins that combine an antibody with a lectin domain to bind glycan motifs and block glycan-driven immune suppression, aiming to restore anti-tumor immunity.

Conventional antibodies excel at targeting proteins. Glycans are trickier: they’re structurally diverse, conformationally flexible, and often shared across healthy tissues. AbLecs attempt to solve this by combining two capabilities in one construct:

1. Antibody-derived targeting/format advantages (familiar scaffolds, Fc engineering options, established manufacturing playbooks)
2. Lectin-based glycan recognition (binding sugar motifs that drive immune-inhibitory signaling)

The Nature Biotechnology briefing describes AbLecs as outperforming conventional antibody therapies—including approved cancer drugs—in both cell-based and animal studies. Even without full details in the preview, the signal is clear: blocking glycan interactions can materially improve immune control of tumors.

Why chimeric design matters

Most companies get glycan targeting wrong by treating it like “just another antigen.” It’s not. Glycans often behave more like patterns than discrete targets.

A chimeric antibody–lectin format can:

• Increase functional avidity through multivalent interactions
• Tune binding to contextual density (tumor surface glycan clustering)
• Potentially reduce dependence on a single protein epitope that tumors can downregulate

This matters because tumors are good at antigen escape. Glycan remodeling is also an escape route—but it’s constrained by cell biology, which can make it a more stable vulnerability.

Where this fits among CD47, Siglecs, and the “don’t eat me” network

Answer first: AbLecs should be viewed as part of a broader effort to disable tumor “don’t attack me” signals, including protein checkpoints (PD-1) and innate checkpoints like CD47 and Siglecs.

If you follow immuno-oncology closely, you’ve seen increasing attention on innate checkpoints. CD47 is the best-known “don’t eat me” pathway, but it’s not the only one. Siglec pathways—especially those involving sialic acids—act like a parallel immune-suppressive lane.

A practical way to think about the landscape:

• PD-1/PD-L1: primarily affects exhausted T cells and adaptive responses
• CD47–SIRPα: blocks macrophage phagocytosis
• Siglec–sialic acid axes: suppress innate immune activation and shape antigen presentation

The attractive thing about glycan checkpoint blockade is combination logic. If a tumor uses both CD47 and sialylation to suppress macrophages, you can get non-redundant synergy by blocking both—something prior research in the space has suggested.

For drug discovery leaders, the question becomes less “Is glycan targeting real?” and more:

“Which patients are suppressed by glyco-immune checkpoints, and can we identify them early enough to treat effectively?”

That’s a biomarker and modeling problem—again pointing directly at AI.

How AI can accelerate AbLec discovery (and de-risk it)

Answer first: AI can speed AbLec development by predicting glycan–protein interactions, optimizing binding specificity, designing safer multispecific formats, and improving patient selection through glycomics-informed biomarkers.

Glycans are notoriously difficult to work with because the design space is huge and the data is fragmented across glycomics, proteomics, structural biology, and immunology. That’s exactly where modern AI methods help—if the program is set up thoughtfully.

1) Modeling glycan interactions: structure, flexibility, context

Lectin–glycan binding isn’t a simple lock-and-key. Glycans are flexible, and small changes in branching can change binding.

AI can contribute in three concrete ways:

• Structure prediction for engineered binders: using protein structure modeling to explore lectin variants and interface geometry
• Binding prediction with uncertainty: ranking glycan motif affinity while quantifying confidence (crucial when experimental glycan arrays are incomplete)
• Context-aware scoring: integrating cell-surface glycan density and presentation rather than assuming free glycans in solution

If your team has tried to optimize binding with only a handful of glycan array readouts, you’ve seen how quickly this becomes a lab bottleneck. Good models don’t replace experiments—but they reduce the number of experiments you need.

2) Multi-objective optimization: potency, specificity, and developability

AbLecs will live or die by tradeoffs:

• Strong enough binding to block suppression in tumors
• Specific enough to avoid broad immune perturbation in healthy tissue
• Developable enough (stability, aggregation risk, manufacturability) to reach clinic

This is a classic multi-objective optimization problem. In practice, I’ve found teams move faster when they treat developability as a first-class target from day one—not a late-stage “we’ll fix it later” activity.

AI-supported workflows can:

• Predict aggregation and viscosity risk from sequence and structure
• Suggest mutations that improve stability while maintaining binding
• Explore Fc designs to tune effector function (when desired) without overshooting safety margins

3) Patient selection: turning glycomics into actionable biomarkers

Glycan targeting without patient selection is a recipe for noisy trials.

A realistic biomarker strategy for AbLecs will likely combine:

• Tumor glycomic signatures (sialylation patterns, glycan density proxies)
• Immune receptor expression (Siglec family expression on myeloid/NK populations)
• Spatial context (where suppressive macrophages sit relative to tumor cells)

AI helps by learning from multimodal data—histology, single-cell, mass spec–derived glycomics, and clinical outcomes—to identify who’s most likely to respond.

If you’re building leads, this is also where value concentrates: enabling sponsors to run smaller, cleaner, faster trials.

What pharma and biotech teams should do next

Answer first: Treat AbLecs as a platform opportunity, but plan for glycan-specific risks—on-target/off-tumor binding, biomarker complexity, and immune-system-wide effects.

AbLecs are exciting, but they’re not “plug-and-play antibodies.” If you’re evaluating this space—either as a therapeutic developer or a partner providing AI drug discovery capabilities—these are the moves that separate serious programs from science projects.

A practical development checklist

1. Define the suppressive mechanism you’re blocking. Is it Siglec-7/9 on NK cells? Siglec-9 on macrophages? Galectin networks? Be explicit.
2. Map on-target/off-tumor risk early. Glycans appear on healthy tissues; what matters is motif, density, and accessibility. Build a tissue cross-reactivity plan from the start.
3. Use combination hypotheses that are biologically justified. Pairing with PD-1, CD47, or macrophage reprogramming agents should be based on tumor microenvironment data, not trend-following.
4. Invest in glycan measurement infrastructure. If you can’t measure the target well, you can’t optimize it well. Partnering for glycomics is often faster than building it all in-house.
5. Make AI useful, not decorative. Tie models to experimental decision points: which variants to build, which cell systems to test, which patient subsets to enroll.

“People also ask” (and the answers you can use internally)

Are glycans druggable targets? Yes, but they’re druggable as patterns and pathways, not as single immutable epitopes.

Why not just target Siglecs with antibodies? You can, and some programs do. AbLecs introduce a different control point: blocking the glycan ligand side and potentially reshaping multiple inhibitory interactions at once.

What’s the biggest risk? Specificity and systemic immune effects. If the binding motif is too common, you may see broader immune modulation than you intended.

The bigger picture for AI in drug discovery

Glycan checkpoint blockade is a reminder that biology’s most important “interfaces” aren’t always protein-protein. They’re often messy surfaces—glycans, lipids, spatial organization, and cell-cell contact rules.

AbLecs are a strong example of where AI in pharmaceuticals & drug discovery can move beyond productivity gains and into strategic advantage: designing better binders, predicting safety earlier, and selecting patients more precisely. If your organization wants to generate leads and partnerships in this space, focus less on generic “AI for immunotherapy” messaging and more on tangible deliverables—binder optimization cycles, developability prediction, and biomarker-driven trial design.

The forward-looking question I’d bet on for 2026 planning is simple: Which tumors are using glycan checkpoints as their primary immune shield—and how fast can we identify them and block them safely?