Examples
Example queries and generated hypotheses.
Real queries across synbio subfields, with representative outputs. The actual generations are longer and include full paper citations, experimental outlines, and stress-test options. These are condensed for browsing.
CRISPR base editing
Could ABE8e off-target activity at bystander A bases be predicted from local RNA secondary structure of the protospacer-flanking sequence?
Representative output
Hypothesis: Bystander A-to-G conversion rates at positions 4–8 of the protospacer scale inversely with the predicted minimum free energy of a 30-nt window centered on the edit site, because secondary structure restricts ssDNA exposure during R-loop formation. Confidence: MEDIUM. Key risk: most published ABE8e characterization assays don't report local RNA-fold metrics, so a retrospective re-analysis of existing deep-sequencing datasets would be required before any prospective test.
A typical LitHypo hypothesis names a measurable quantity, ties it to a mechanistic claim, and surfaces the practical risk that determines whether the experiment is feasible. The MEDIUM confidence reflects strong indirect literature support but no direct prior test.
Protein engineering
Can rationally engineered allosteric switches in TEV protease be designed using coevolution-derived contact networks?
Representative output
Hypothesis: Insertion of a small-molecule binding domain (e.g., FRB) at positions identified as coevolutionary hubs by EVcouplings or DCA in the TEV protease catalytic core will yield rapamycin-responsive activity changes greater than insertions at non-hub positions. Confidence: MEDIUM-HIGH. Supporting papers: Reynolds et al. 2011 sector analysis; Diss & Lehner 2018 epistasis maps; recent allosteric switch designs in dihydrofolate reductase.
Cross-disciplinary queries — coevolution methods plus protein engineering — are where LitHypo most clearly outperforms general chatbots, which tend to retrieve only the most-cited papers on each side.
Biomolecular condensates
What controls the difference in viscosity between FUS and TDP-43 condensates in vitro?
Representative output
Hypothesis: Apparent condensate viscosity differences between FUS and TDP-43 at matched protein concentration are driven primarily by aromatic residue density in the disordered region, not by RNA content. Confidence: MEDIUM. Identified risks: most published rheology measurements differ in salt and crowding conditions, so a controlled comparison would require new experiments rather than literature meta-analysis.
The engine deliberately flags when the existing literature cannot resolve the hypothesis without new experiments — important context the researcher needs before designing follow-up work.
Plant synthetic biology
Can synthetic CRE-loxP-based memory circuits record drought stress history in Arabidopsis root cells?
Representative output
Hypothesis: A drought-responsive promoter (RD29A) driving Cre recombinase, paired with a constitutively expressed loxP-stop-loxP reporter, will permanently mark cells that experienced osmotic stress at least once during root development, with single-cell resolution. Confidence: HIGH. Key risk: leakiness of RD29A under non-stress conditions has been reported and would create false-positive recording.
HIGH confidence is reserved for cases where every component of the proposed system has been individually validated in the published literature.
Organoid systems
Can engineered Notch signaling perturbations rescue patterning defects in DLL3-deficient cerebellar organoids?
Representative output
Hypothesis: Optogenetic activation of Notch1 ICD release (via LightOn-Cre + floxed NICD) at 24-hour intervals during days 14–28 of cerebellar organoid differentiation will partially rescue rhombic lip patterning in DLL3 knockout backgrounds. Confidence: MEDIUM. Key risk: pulsatile Notch signaling fidelity in 3D organoid contexts has been characterized far less than in 2D cultures.
Organoid queries surface a recurring tension: most mechanistic work has been done in monolayer culture, so any 3D hypothesis carries a translation risk the engine surfaces explicitly.
Metabolic engineering
Can a synthetic anti-CRISPR-based feedback loop stabilize CRISPR interference of essential metabolic genes at sub-lethal levels?
Representative output
Hypothesis: Coupling dCas9-mediated knockdown of a fitness-essential gene (e.g., E. coli folA) to a metabolite-responsive expression of AcrIIA4 (anti-CRISPR) will yield homeostatic knockdown depth that self-tunes to growth rate, enabling sustained partial knockdown of essential genes for metabolic flux studies. Confidence: MEDIUM. Supporting papers: recent AcrIIA characterization work; published anti-CRISPR-based feedback circuits.
The engine often proposes circuit-level designs that combine published parts in untested configurations — exactly the kind of hypothesis worth pressure-testing before bench work.
Gene circuits
What's the smallest synthetic circuit that produces robust oscillations in mammalian cells without manual induction?
Representative output
Hypothesis: A two-component negative-feedback loop using TetR-VPR self-repression with a destabilized fluorescent reporter and short half-life mRNA UTRs will produce sustained oscillations with period 6–12h in HEK293 cells, more robustly than the canonical three-node repressilator topology. Confidence: MEDIUM-LOW. Key risk: published mammalian repressilator implementations report substantial cell-to-cell variability that may not be resolved by topology simplification.
MEDIUM-LOW confidence rulings push the user toward computational pre-screening (e.g., stochastic simulation) before committing to construction.
Synthetic transcription factors
Can zinc finger-based synthetic TFs achieve better silencing specificity than dCas9-KRAB at endogenous promoters?
Representative output
Hypothesis: Engineered zinc finger arrays fused to KRAB will achieve narrower off-target silencing profiles than dCas9-KRAB at the same target genes in K562 cells, because the smaller DNA footprint of zinc fingers reduces incidental binding at homologous sites within 50 bp. Confidence: MEDIUM. Risks: published direct head-to-head comparisons at matched epigenetic backgrounds are limited, so meta-analysis across separate datasets carries platform-effect confounds.
When direct comparisons don't exist in the literature, the engine surfaces that gap rather than synthesizing across non-comparable studies.
Want to understand the engine? Read the methodology →
Want concrete workflows? See use cases →