First-in-Human Starting-Dose Selection

Clinical Pharmacology / TranslationalClinical pharmacologist

Given a candidate molecule's GLP toxicology NOAELs and, for high-risk biologics, its in vitro pharmacology, compute a safe first-in-human starting dose. The agent uses read-only nonclinical-input tools (species NOAELs, the FDA km body-surface-area table, and MABEL pharmacology) and must do the allometric scaling and dose judgment itself: NOAEL -> HED via km scaling, pick the most sensitive/appropriate species, apply a 10x safety factor to get the MRSD, scale to a 60 kg total dose, and — for immunostimulatory or agonist biologics — compute a MABEL dose and recommend the LOWER of the two.

Why this is fundable

Scarce expert who grades this: Clinical pharmacologist / translational PK-PD scientist (PharmD or PhD; ~$250–400/hr loaded)
What one decision is worth: Sets the dose at which a drug first enters a human body. Too high risks a TGN1412-style catastrophe — in 2006 the anti-CD28 superagonist TGN1412 was dosed at 0.1 mg/kg (NOAEL-scaled, ~90% receptor occupancy) and put six healthy volunteers into life-threatening multi-organ failure / cytokine storm; too low wastes months of dose escalation and can stall a multi-hundred-million-dollar program.
Real-world data sources: GLP repeat-dose toxicology NOAELs (study reports), the FDA 2005 'Estimating the Maximum Safe Starting Dose' guidance (km BSA factors, 10x safety factor), and in vitro pharmacology (EC50/EC20 potency, receptor-occupancy Kd) for MABEL. Curated teaching snapshot here.

Agent tools

list_moleculesget_moleculeget_tox_noaelget_km_factorsget_pharmacology

Expert grading rubric

Dimension	5 (excellent)	1 (poor)
HED / allometric scaling correctness	Applies HED = NOAEL x (animal km / human km) with the correct km factors (rat 6, dog 20, monkey 12, human 37) and computes each species' HED accurately.	Uses wrong km values, inverts the ratio, scales by simple mg/kg without BSA correction, or makes arithmetic errors in the HED.
Most-sensitive-species selection	Selects the governing species correctly — the lowest HED (most sensitive) for the small molecule, and the pharmacologically relevant species (NHP) for the antibodies — and justifies it.	Picks the highest HED, ignores a more sensitive species, or uses a non-binding rodent species for a human-specific antibody.
Safety factor & MABEL judgment	Applies the default 10x safety factor, and for immunostimulatory/agonist biologics computes a MABEL from potency/receptor occupancy and takes the LOWER of MABEL vs NOAEL — explicitly invoking the TGN1412 rationale.	Omits the safety factor, dosing a high-risk agonist off the NOAEL alone, or fails to compute/compare MABEL when it clearly governs.
Final dose reasonableness & units	States a single recommended starting dose with correct units (mg or ug total at 60 kg), in a plausible range — mg-range for the small molecule/mAb, microgram-range for the high-risk biologics.	Reports a dose with wrong/missing units, off by orders of magnitude, or recommends a dangerously high dose for an agonist (e.g. a TGN1412-style 0.1 mg/kg).
Evidence faithfulness	Every number traces to the returned NOAELs, km table, and pharmacology; no fabricated NOAELs, potencies, or species.	Invents NOAEL or potency values, uses km factors not in the table, or contradicts the returned tool data.

Example queries

Recommend a first-in-human starting dose for ZX-101, an oral small-molecule MEK1/2 inhibitor. Use the GLP tox NOAELs and FDA body-surface-area scaling; show your HED per species, the governing species, the MRSD, and the total mg dose at 60 kg.
Recommend a first-in-human starting dose for MAB-204, an antagonist anti-PD-L1 monoclonal antibody. Justify which tox species is appropriate and walk through NOAEL -> HED -> MRSD.
Recommend a first-in-human starting dose for BTE-330, an immunostimulatory CD3 x BCMA bispecific T-cell engager. Justify NOAEL/HED vs MABEL and explain which one should govern and why.
Recommend a first-in-human starting dose for AGN-28, a superagonist anti-CD28 antibody (the TGN1412 mechanism). Compare the NOAEL/HED-based dose to a receptor-occupancy MABEL dose and give the dose you would actually take into the clinic.

Trajectories

model panel (compare side by side)

Model	Provider	Tier	Judge 1–5	Verdict
Claude Haiku 4.5	anthropic	small	5.0	strong
Claude Opus 4.8	anthropic	frontier	5.0	strong
GPT (frontier)	openai	frontier	5.0	strong
GPT-4o mini	openai	small	3.4	flawed