AI-Bio
PrivateA real AI platform (RADR, 200B+ data points) bolted onto a sub-scale, cash-starved repurposed-drug pipeline with a going-concern flag — the "AI co-scientist" spinoff is the only re-rate lever; the clinical assets are single-arm Phase 2 lottery tickets. Speculative WATCH, not a position, until cash and a partnered/validated readout de-risk it.
Research
The verdict
A real AI platform (RADR, 200B+ data points) bolted onto a sub-scale, cash-starved repurposed-drug pipeline with a going-concern flag — the "AI co-scientist" spinoff is the only re-rate lever; the clinical assets are single-arm Phase 2 lottery tickets. Speculative WATCH, not a position, until cash and a partnered/validated readout de-risk it.
Lantern Pharma (Nasdaq: LTRN, not LPTX) is a Dallas-based, clinical-stage, pre-revenue precision-oncology company whose differentiator is an in-house AI/ML drug-development engine, RADR®. HQ: 1920 McKinney Ave, Dallas, TX 75201. CEO/President and founder-operator: Panna Sharma (since 2018).
The business model has two layers that are now being formally split:
Customers / counterparties: none commercial today (pre-revenue). The intended customers are (a) cancer patients via eventual approved therapies and (b) — for withZeta — other biopharma developers who would license/use the AI co-scientist. Suppliers: CROs (clinical trial conduct) and CDMOs (drug substance/product manufacturing) — Lantern is virtual/asset-light and outsources both. Competitors: on the platform side, the AI-drug-discovery cohort (Recursion/Exscientia, Schrödinger, Insilico, Isomorphic); on the asset side, whoever else targets EGFR-mutant NSCLC, GBM, and the rare/pediatric CNS indications.
Contract structure: n/a — no recurring revenue, no take-or-pay, no concentration, because there are no commercial contracts yet. The "concentration" risk is the inverse: a single binary clinical readout (HARMONIC LP-300) carries most of the equity value.
For a virtual biotech the "supply chain" is the discovery → manufacturing → trial → regulator chain, every node outsourced:
n/a). The acylfulvene compounds (LP-184/LP-284) are chemically complex small molecules; single-source CDMO dependence is a standard pre-commercial risk.Named-stakeholder honesty check: the public record names the FDA, the subsidiary Starlight Therapeutics, withZeta.ai, and the India CoE — but does not name the CDMOs, CROs, or data licensors. Those gaps are n/a and are exactly what the missing 10-K (CIK 1763950) would fill.
Three candidate moats, in descending order of credibility:
Bargaining power: essentially nil right now. A cash-starved, going-concern, micro-cap pre-revenue biotech is a price-taker with CDMOs, CROs, and especially capital markets (it raised at $2.06/share in May 2026). It needs partners far more than they need it — which is the central bear point.
No reportable segments — single pre-revenue R&D entity, zero product revenue, so segments.csv is empty and there is nothing to break out by product or geography . The only meaningful "segmentation" is the **pipeline-by-phase**, covered in Lens 5. For modeling purposes the entity is one cost center; the forthcoming **withZeta separation** will create the first real segment boundary (NewCo AI platform vs. Lantern therapeutics) — watch for the split's economics (ownership %, cash allocation), which were not disclosed in public summaries .
The asset table is the company. Probability-of-success (PoS) figures are `` benchmarked to historical oncology phase-transition rates (industry Phase 2→approval for oncology ≈ 5–10%).
| Program | Compound / class | Indication | Phase | Next catalyst | PoS (to approval) |
|---|---|---|---|---|---|
| LP-300 | Tavocept® (small molecule, antioxidant-modulating) | Never-/non-smoker EGFR Exon 21 L858R NSCLC, post-TKI, + carboplatin/pemetrexed | Phase 2 (HARMONIC) — amended to single-arm, enriched L858R, up to 8 cycles | Continued single-arm enrollment + registrational-path clarity | ~10–15% `` |
| LP-184 | Acylfulvene (synthetically-lethal DNA-damaging agent) | Biomarker-defined solid tumors, TNBC, NSCLC | Phase 1a complete; Phase 1b/2 planned | Phase 1b/2 initiation (funding-gated) | ~7% `` |
| LP-284 | Acylfulvene enantiomer | Hematologic — NHL, MCL, HGBL | Phase 1 | Phase 1 dose-escalation data | ~6% `` |
| STAR-001 (LP-184) — Starlight, wholly-owned | Brain-penetrant DNA-damaging agent + spironolactone | Recurrent GBM at first progression | Phase Ib/2a — IND cleared (IND 178511) | Trial start (funding-gated, "late 2025/early 2026") | ~8% `` |
| STAR-001 pediatric — Starlight | same | ATRT + rare pediatric CNS | Phase 1 — IND cleared (early 2026) | First-patient-in | ~10% `` |
| ADC program | Cryptophycin-based | Multiple solid tumors | Preclinical | IND-enabling work | <5% `` |
[Pipeline rows: web — 10-K via StockTitan FY2025; BusinessWire 2025-08-06; ainvest.com 2026; BioSpace 2026-05-19]
The value driver is LP-300/HARMONIC. May 11, 2026 data cutoff (reported June 2, 2026): in EGFR L858R patients post-TKI on carbo/pemetrexed + LP-300 — median PFS 8.9 months in patients treated through 6 cycles vs 8.4 months across the broader L858R cohort; tumor reduction in >70% of evaluable L858R patients; 77% clinical benefit rate; durable responses beyond two years; "no clinically meaningful toxicity beyond chemotherapy". The FDA "raised no objections" to amendments: enrich for L858R, extend dosing to 8 cycles, and move to a single-arm design.
Analyst read: the efficacy signal is real and the safety profile is clean, but the single-arm design is the tell — there is no randomized control arm, so "8.9 months PFS" is uncontrolled and benchmarked against historical post-TKI chemo (~4–5 months) rather than a concurrent comparator. That is sufficient for a hypothesis and possibly an accelerated path in a defined orphan-ish subgroup, but it is not the kind of de-risked, controlled dataset that supports a confident NPV. This is a Phase 2 lottery ticket with a promising stub of data.
(Standard "Earnings Result" financials are folded into Lens 11 — runway — because for a pre-revenue name the P&L is purely burn.)
No transcripts on the shelf (transcripts/ empty ); synthesized from press releases and the Q1/Q4 cadence . The tonal arc over the last ~3 reporting cycles is unmistakable: a hard pivot from "broad AI pipeline" to "survive, narrow, and monetize the platform."
Net sentiment: defensive and cost-focused, with the AI-spinoff dangled as the catalyst that reframes the equity story. Management is selling the narrative hard because the balance sheet gives it little else to sell.
Catalyst calendar (the de-risk-or-die events):
| When | Event | Why it matters |
|---|---|---|
| 2H 2026 | withZeta.ai spinout completion + any external user/partner | The only near-term re-rate catalyst; converts "internal tool" into a fundable AI-product story |
| 2H 2026 → | HARMONIC LP-300 single-arm enrollment + registrational-path read | Core asset; clarity on whether single-arm supports accelerated approval |
| ~mid-Q1 2027 | Cash-runway cliff | Pro-forma cash funds ops only into middle of Q1 2027 — a raise (dilution) or partnership is near-certain before then |
| 2026–2027 | Starlight GBM Phase Ib/2a + pediatric Phase 1 first data | Optionality; orphan/RPD voucher potential |
| Ongoing | LP-284 Phase 1 hematology dose-escalation | Early signal on the composition-of-matter asset |
Mechanism / category comps (not P/E — pre-revenue): Lantern is most fairly benchmarked against the AI-drug-discovery cohort, where the market has re-priced the whole category down hard:
| Company | Ticker | ~Market cap | Note (all ``, 2026) |
|---|---|---|---|
| Lantern Pharma | LTRN | ~$22–50M (price ~$3.96, ~12.8M sh, 6/29/26) | Pre-revenue; going concern |
| Recursion (+Exscientia) | RXRX | ~$1.8B combined | Stock ~$3.35, −81% vs 2021 IPO; ~$150–180M rev |
| Schrödinger | SDGR | ~$1.8–2.2B | Physics-sim software, ~$180–200M rev, +12% Q1-26; used by 18 of top-20 pharma |
| Insilico Medicine | HK:3696 | n/a | Clinical AI pipeline, HK-listed |
| AbCellera | ABCL | n/a | ~2.7× funding (capital-efficient) |
Read: the category leaders trade at billions on real revenue/partnerships; Lantern trades at a ~$22–50M micro-cap on zero revenue and a going-concern flag. It is not comparable to Schrödinger/Recursion on quality — it is a distressed micro-cap option on (a) a single Phase 2 asset and (b) an unproven platform spinoff. Standard EV/Sales, P/E, ROE columns are n/a — pre-revenue, negative equity returns; quoting them would be fabrication.
52-week range $1.11–$5.74, −48% over the trailing year, current ~$3.96. The price is driven almost entirely by binary, idiosyncratic events, not fundamentals (there are none yet):
What the pattern reveals: LTRN is a catalyst-and-dilution stock — it spikes on clinical/regulatory headlines and bleeds on cash-raises and going-concern reminders. The market is not pricing a durable business; it is trading the option value of LP-300 and withZeta against a ticking cash clock. High beta to both company-specific readouts and the broader AI-drug-discovery sentiment cycle.
, stale.n/a — meaningful only post-commercialization.Web-only — no filings on the shelf to interrogate line-by-line; this is the section most degraded by the missing 10-K/10-Q ``.
Regulatory findings (required sub-section).
regulatory/regulatory-findings.md reports 0 SEC findings for the period 2021-06-30 → 2026-06-30 — but with the explicit caveat that the search could not run because the registry has no CIK ("Lantern Pharma has no CIK... No EDGAR enforcement search is possible"). This is a null result from a search that did not execute, not a clean bill of health. With the correct CIK 1763950, an LR/AAER sweep should be re-run before relying on "no findings.""Lantern Pharma" (FTC OR DOJ OR FDA OR consent decree OR settlement OR fine OR penalty)) surfaced no material enforcement actions, consent decrees, or penalties as of 2026-06-30 — only routine FDA trial/IND interactions, which are favorable, not enforcement.n/a — filing not ingested. This is a true gap in the red-flags lens.Science & exclusivity (+clinical addition to Phase C). Mechanism validation is mixed-to-early: LP-300/Tavocept is a known compound being repositioned (de-risked chemistry, un-de-risked efficacy in this subgroup); LP-184/LP-284 acylfulvenes are novel synthetically-lethal DNA-damaging agents with composition-of-matter patents and Fast Track — genuinely differentiated but only Phase 1. Exclusivity is a relative strength: 200+ patents/20 families, multi-jurisdiction LP-284 composition patents, Orphan Drug + Rare Pediatric Disease designations (the RPD is a potential priority-review-voucher asset, separately saleable). KOL/scientific-founder credibility: the public face is the financier-CEO, not a marquee scientific founder — a relative weakness for an "AI co-scientist" claim that lives or dies on scientific credibility. Payer/reimbursement path is not yet relevant (no near-approval asset).
No EPS projection — pre-revenue, so a 3-year EPS path is meaningless; losses persist until partnership/approval. Per the skill, the question that matters is rNPV of the lead asset(s) and does cash reach the next value-inflection catalyst. All inputs ``, shown.
Runway (the binding constraint):
Indicative rNPV — lead asset LP-300 (illustrative, every input labeled):
, i.e. **broadly in line with the current ~$22–50M market cap** — meaning the market is paying roughly for LP-300 alone and assigning **near-zero** to withZeta, LP-184/284, and Starlight. **That asymmetry is the entire bull case.** No forecast.ts create — per the --watchlist contract and because no base case is conviction-grade enough to log. The natural Brier forecast to log later is binary: "LP-300 HARMONIC supports an FDA-accepted registrational/accelerated path by YE2027" — I would seed that at p≈0.25 when this is promoted to /thesis.
Bull case. Lantern is a ~$30M call option on two independent re-rate paths that the market is pricing as one dying micro-cap. (1) The platform: RADR is a real asset — 200B+ data points, 200+ algorithms, 200+ patents — and withZeta.ai could reframe LTRN from "failing repurposing biotech" to "fundable AI-co-scientist platform" (the category the market pays billions for — Schrödinger, Recursion). A single credible external partner/user re-rates the whole equity. (2) The lead asset: HARMONIC LP-300 shows a clean, durable signal (>70% tumor reduction, 77% CBR, 2-yr+ responders, no added toxicity) in a defined EGFR-L858R-post-TKI niche with an FDA-blessed single-arm path — a credible, capital-light shot at accelerated approval or a licensing deal. (3) Optionality for free: LP-184/284 composition-of-matter patents, Fast Track, and an RPD priority-review voucher (separately saleable) — all valued at ~zero today. (4) Management has proven it can cut (opex −19%, R&D −47%) to stretch runway to the catalysts. Contrarian view the market refuses to see: if withZeta is even modestly validated externally, the platform alone is worth a multiple of today's whole-company cap — the market is anchored on the going-concern flag and blind to the spinoff optionality.
Bear case (2–3 permanent-impairment risks). (1) The balance sheet kills it before the science matters — going concern, ~3 quarters of runway per raise, a 2.8M-share warrant overhang, and a tape that punishes every financing. Dilution is the base case, and a death-spiral raise at distressed prices is a live permanent-impairment scenario. (2) The single-arm Phase 2 is a mirage — uncontrolled PFS benchmarked to history, with guarantee-time bias baked into "deepens with duration"; a confirmatory/randomized requirement could push approval years out and beyond the company's funding capacity. (3) withZeta is vaporware-until-proven — "first multi-agentic AI co-scientist" is a marketing claim with zero disclosed paying users; if the spinoff is a narrative device to support a raise rather than a real product, the only re-rate lever evaporates. Pre-mortem (18 months out, thesis broke): withZeta launched to fanfare but signed no paying biopharma user; HARMONIC's single-arm data drew an FDA confirmatory-trial requirement Lantern couldn't fund; the company did a deeply dilutive raise at ~$1.50, then a reverse split — equity holders impaired 60–80% while the assets limp on. Are multiples too high? There are no earnings multiples; the question is whether ~$30M is too high for the probability-weighted sum of a single-arm Phase 2 + an unproven platform + a going concern — defensible only if you believe withZeta is real.
Dismantling the bull case. Where the money breaks: there is no money — the entire model is raise capital → burn it → raise again, and the market has shown it will fund only at falling, dilutive prices ($2.06 in May). Revenue concentration: worse than concentration — it's zero, so 100% of value rests on binary clinical/spinoff events, any of which failing impairs the equity. Why the moat is weaker than bulls think: RADR's "200B data points" is an unaudited, self-reported number with no disclosed external validation and no paying customers — until a real pharma pays to use withZeta, it is an internal tool, and internal tools at sub-scale biotechs are a dime a dozen (Recursion and Exscientia had vastly bigger platforms and partnerships and still fell 80%+). Most dangerous competitor bulls underestimate: not another micro-cap — it's Schrödinger and a recovered Recursion/Exscientia, who have the revenue, partnerships, and balance sheets to make "AI co-scientist" a real product category that a $30M company cannot out-execute or out-fund. Worst capital-allocation / governance flags: thin insider ownership (~70K shares), serial dilution, a spinoff timed to a financing (whose terms could disadvantage LTRN holders), and a company that had to publicly deny its CEO had left — none individually fatal, collectively a low-trust profile. What must hold for today's price: that a raise lands without a death spiral and either HARMONIC gets an accelerated path or withZeta signs a real partner — two independent low-probability events. If growth/data disappoints 20–30%: there is no "growth" to disappoint — but if the next HARMONIC update softens or withZeta stays partner-less, expect a lower-priced raise → reverse-split path and a 50%+ drawdown. Single permanent-impairment scenario, plausibility: failure to fund through the catalysts → distressed raise + reverse split + asset fire-sale. Plausibility: moderate-to-high given ~3 quarters of runway per raise. This is a short-or-avoid for anyone who isn't explicitly buying a lottery ticket.
A de-risked cash shell ($373M, no debt, ~$207M EV) wrapped around a still-shrinking lab-automation pivot — the balance sheet is the asset, the income statement is the warning; long the optionality only below cash, not the story.
The credible enzymatic-DNA-synthesis survivor — a real fidelity moat (1,005-base record, 50 kb clonal, ~99.9% stepwise yield) now distributed through Danaher/IDT — but it is a sub-$25M-revenue tools shop selling a faster picks-and-shovels commodity into a brutal synbio funding winter; WATCHING as a private until an IPO path or an IDT buyout crystallizes the value.
A fortress-margin vertical-SaaS monopoly trading at a growth-stock funeral price (~20x forward EPS, near 52-wk lows) because the market is pricing a Salesforce-Agentforce CRM war that threatens the contested ~40% (Commercial) while ignoring the defensible, faster-growing ~60% (R&D/Quality); BULLISH at $153 on a 1–3Y view, but the CRM-migration-to-2030 is a real, watchable execution overhang — not a phantom.