Reading Normal-Tissue Expression for Radioligand Dosimetry Risk

In radioligand therapy, the thing that hurts the patient and the thing that helps them are the same mechanism pointed at different tissue. The radioligand delivers a dose wherever the target is expressed. In tumor, that's therapy. In a healthy kidney or a salivary gland, that's toxicity. So one of the first questions in evaluating any radioligand target is: where else, in normal tissue, does this protein show up — and how much does that matter?

Normal-tissue expression data is the cheapest, fastest way to answer the first half of that question during early triage. It's also routinely over-read. This piece covers how to use it well: which organs matter, how to interpret the numbers, and — importantly — the point where expression data stops being predictive and real dosimetry has to take over.

Triage signal, not a dose estimate. Everything below helps you rank and screen targets before committing resources. None of it is a substitute for measured biodistribution and clinical dosimetry. If you take one thing away, make it the distinction in the last section.

The dose-limiting organs in radioligand therapy

Not all normal tissue carries equal risk. A few organs dominate radioligand toxicity, either because they're radiosensitive or because they naturally accumulate radioligands regardless of target biology:

• Kidney. The classic dose-limiting organ for peptide radioligands. Small peptides are cleared renally and reabsorbed in the proximal tubules, concentrating dose there. High renal target expression compounds a risk that clearance already creates.
• Bone marrow. Highly radiosensitive, and the usual constraint for potent alpha emitters like ²²⁵Ac, where even low off-target dose to marrow drives hematologic toxicity.
• Salivary glands. The source of the xerostomia (dry mouth) that limits ¹⁷⁷Lu-PSMA dosing — a direct, well-documented case of on-target, off-tumor expression setting the therapeutic ceiling.
• Liver. High perfusion and a major clearance route; relevant for hepatic uptake and background.

When you screen a target's normal-tissue profile, these are the organs to look at first. High expression in any of them is a yellow-to-red flag that has to be weighed against tumor uptake.

What expression data can — and can't — tell you

Two caveats determine how much weight to put on the numbers.

RNA versus protein. Most large, genome-wide normal-tissue atlases are transcriptomic — they measure mRNA (e.g. nTPM), not protein. mRNA correlates with protein imperfectly; a gene can be transcribed without much surface protein, and vice versa. Transcript-level data is an excellent first filter precisely because it's comprehensive and cheap, but protein-level evidence (immunohistochemistry, proteomics) is closer to what the radioligand actually sees. Where protein data exists, prefer it; where it doesn't, treat RNA as a screening signal, not a conclusion.

Expression is not absorbed dose. This is the big one. The absorbed dose to an organ depends on the radioligand's pharmacokinetics — how fast it clears, by which route, how long it dwells — as much as on target expression. The kidney is the sharpest example: renal dose in peptide radioligand therapy is driven heavily by tubular reabsorption and clearance kinetics, which no expression database captures. A target with low kidney expression can still deliver meaningful renal dose if the ligand is retained there. Low normal-tissue expression is a favorable prior, not a safety result.

Used with those two caveats in mind, expression data does real work.

Reading the Human Protein Atlas for target triage

The Human Protein Atlas (HPA) is the practical workhorse for this. Its consensus RNA dataset gives normalized expression (nTPM) across normal tissues, and its pathology/IHC data adds protein-level and cancer expression. For triage against the dose-limiting organs, simple bands are enough to separate candidates:

Apply these to kidney, liver, bone marrow, and salivary gland, and you can quickly separate targets with a clean normal-tissue profile from those carrying built-in liabilities. It won't rank the survivors precisely — that's what tumor specificity and, eventually, dosimetry do — but it reliably screens out the obviously risky ones.

(A worked note: PSMA itself would fail a naïve "low everywhere" screen — it's expressed in renal tubules and salivary glands, which is exactly why xerostomia and renal uptake are its signature toxicities. It succeeds anyway because tumor uptake and internalization are extraordinary. That's the whole point: normal-tissue expression is one criterion among six, not a veto.)

Tumor-to-kidney ratio as a first-pass safety signal

Because the kidney is so often the dose-limiting organ, the tumor-to-kidney expression ratio is a useful single number for early screening — it folds efficacy signal (tumor expression) and the dominant safety constraint (renal expression) into one comparison. Treat it as continuous, not pass/fail: a target with a 2,000× tumor-to-kidney ratio is genuinely more attractive than one at 30×, even though both look "good" against a simple threshold.

Two guardrails:

• Pair it with patient positivity. A high ratio in a target expressed by only a small fraction of patients is a narrow opportunity, not a broad one.
• Remember what it is. It's an expression ratio, not a dose ratio. It ranks candidates; it doesn't predict the therapeutic index.

Where triage ends and dosimetry begins

Here's the honest handoff. Normal-tissue expression, protein data, tumor-to-background ratios — all of it is triage. It tells you which targets are worth the cost of the next step. It cannot tell you the absorbed dose to the kidney, the therapeutic index, or whether a specific ligand will clear cleanly. Those answers come from measured biodistribution, imaging, and clinical dosimetry with the actual radioligand — the expensive, slow, indispensable work that expression data exists to protect.

A tool that blurs that line does scientists a disservice. The value of expression-based triage is precisely that it's fast and cheap enough to run across thousands of candidates, so the expensive work is spent only on the ones that survive it.

That's how Nuclens uses this data. We pre-compute normal-tissue expression across the dose-limiting organs, tumor-to-background ratios, and patient positivity for roughly 15,000 oncology targets — every value sourced and traceable — so you can screen the field in minutes and reserve the wet lab for the shortlist that earns it.

Screen your targets — free →

Nuclens is a first-pass target-triage platform for radioligand therapy. Expression data indicates risk, not absorbed dose; it does not replace measured biodistribution, clinical dosimetry, or experimental validation.