Sigmatic Scout

Ask a research question.
Get an evidence-backed answer.

Scout orchestrates hundreds of specialized AI agents to run multi-step scientific investigations autonomously. One prompt. Full provenance.

Request Early Access See It In Action

Real Investigations

One prompt. A complete scientific investigation.

These are real Scout outputs from actual research questions run on the platform. No cherry-picking, no cleanup.

Research Question

"For a CAR-T therapy targeting FOLR1 in ovarian cancer, determine the on-target/off-tumor safety profile by profiling expression across healthy tissues, identifying highest-expressing cell populations, assessing overlap with known toxicity patterns, and producing a tissue-level safety risk map."

Investigation Stats

13

Pipeline Turns

3

Data Sources

35

Tissues Profiled

10

PMIDs Cited

Agents Orchestrated

CELLxGENE Census ×3 GTEx Expression (54 tissues) PubMed Search & Fetch Human Protein Atlas Composite Risk Scoring Safety Risk Map Synthesis ×2 Final Definitive Report

Scroll to explore full report ↓

FOLR1 CAR-T On-Target / Off-Tumor Safety Risk Map

Ovarian Cancer Program — IND-Enabling Safety Assessment

Prepared for: Kevin Target: FOLR1 / Folate Receptor Alpha Modality: CAR-T cell therapy Indication: Ovarian cancer

1Original Research Question

For a CAR-T therapy targeting FOLR1 (folate receptor alpha) in ovarian cancer, determine the on-target/off-tumor safety profile by: (1) profiling FOLR1 expression across healthy human tissues and cell types in single-cell atlases; (2) identifying specific cell populations with highest normal-tissue FOLR1 expression; (3) assessing overlap with known CAR-T toxicity patterns from published FOLR1-directed programs; and (4) producing a tissue-level safety risk map ranking organs by predicted FOLR1-mediated toxicity risk.

2Executive Summary

This report consolidates Scout's complete investigation into the on-target/off-tumor safety landscape for a FOLR1-directed CAR-T program in ovarian cancer. Evidence was synthesized from three independent data streams: the CELLxGENE Census single-cell RNA-seq atlas (three successive queries spanning more than 100 cell-type entries across nine tissue categories), GTEx bulk RNA-seq (54 tissues, median TPM), and eight directly relevant published studies identified by targeted PubMed retrieval.

Primary conclusion: FOLR1 is a conditionally actionable but narrowly windowed CAR-T target in ovarian cancer. A validated hematopoietic safety window and near-absent normal ovarian stromal expression define a genuine tumor-selective opportunity. However, five epithelial organ systems carry CRITICAL-tier on-target/off-tumor liability by convergent evidence, and four additional tissues carry HIGH-tier concern.

3Evidence Timeline — Major Pipeline Steps

Step 1–2 · CELLxGENE Census Query 1

Pan-Tissue FOLR1 Expression — Blood, Bone Marrow, Liver, Respiratory, Eye, Brain

First atlas query returned expression data across 6 tissue-general categories and ~100 cell-type entries. Key findings: respiratory club cells (expression 0.593, 63/209 cells = 30.1%) were the highest expressers. All hematopoietic and immune cells in blood and bone marrow showed zero or negligible signal — 33 bone marrow cell types returned expression = 0.0.

Step 3–4 · CELLxGENE Census Query 2

Gap-Tissue Query — Kidney, Brain/Choroid Plexus, Ovary

The renal principal cell is the single highest-expressing normal cell type in the entire combined dataset (expression 2.547; 59.0% of 117 cells). Normal ovarian stromal cells showed expression 0.00023 (3/12,837 cells <0.1%) — confirming tumor-selectivity in the ovarian compartment.

Step 5–6 · PubMed Literature Retrieval

8 Targeted Queries · 8 Directly Relevant Studies Retrieved

FOLR1-TCB bispecific produced target-dependent toxicity in both Lung-Chip and Intestine-Chip platforms (PMID 34378534). Three independent studies confirmed FOLR1 protein at the renal proximal tubule brush-border membrane by IHC. Radiofolate biodistribution documented 18.48 %ID/g renal retention versus 2.33 %ID/g tumor — an 8-fold renal-over-tumor excess.

Step 9–10 · GTEx Bulk RNA-seq (54 Tissues)

Orthogonal Bulk Validation and Gap-Tissue Characterization

Lung at 109.863 TPM was the highest-expressing tissue of all 54 profiled. Minor Salivary Gland (79.135 TPM) and Thyroid (33.956 TPM) emerged as new CRITICAL-tier tissues. Kidney Cortex (50.508 TPM) and Medulla (52.793 TPM) confirmed CRITICAL status.

4Tissue-Level Safety Risk Map

4.1 — CRITICAL Tier: Highest On-Target/Off-Tumor Liability

Tissue	Top Cell Type	Peak Expr (raw)	% Cells	GTEx (TPM)	Composite	Evidence
Kidney	Renal principal cell	2.547	59.0%	50–53	3.005	IHC confirmed ×3; 18.48 %ID/g renal retention; pan-nephron expression
Lung	Club cell (+goblet, epithelial)	0.593	30.1%	109.9	0.268†	Organ-on-chip functional confirmation; GTEx highest of 54 tissues
Salivary Gland	Unresolved — coverage gap	—	—	79.135	GTEx-only	Second-highest GTEx bulk signal; CRITICAL pending validation
Thyroid	Unresolved — coverage gap	—	—	33.956	GTEx-only	Fourth-highest GTEx signal; CRITICAL pending validation
Choroid Plexus	Choroid plexus epithelial cell	0.408	29.0%	~0.3–1.3	0.237†	FRα autoantibody blockade produces CNS folate deficiency; mechanistic override applied

4.2 — Additional Kidney Cell Types (Pan-Nephron Liability)

Cell Type	Expression (raw)	% Cells	Count	Note
Renal principal cell	2.547	59.0%	69/117	Highest expresser in dataset
Columnar/cuboidal epithelial (kidney)	1.140	51.2%	66/129	—
Loop of Henle thick ascending limb	0.975	43.6%	287/658	Reliable estimate (n=658)
Proximal tubule epithelial cell	0.781	41.5%	883/2127	Largest cell count — highest confidence
Distal convoluted tubule epithelial	0.742	38.9%	107/275	—
Collecting duct cell	0.672	39.2%	49/125	—
Renal alpha-intercalated cell	0.396	18.1%	72/397	Lower but non-negligible

4.3 — HIGH Tier

Tissue	Top Cell Type	Peak Expr	% Cells	GTEx (TPM)	Key Evidence
Stomach	Mucous neck cell of gastric gland	1.590	27.8%	10.647	scRNA-seq confirmed; foveolar cells also expressing
Pituitary	Unresolved	—	—	7.113	GTEx signal; mechanistically plausible
Fallopian Tube	Unresolved	—	—	5.845	Relevant to HGSOC cell of origin
Intestine	Unresolved	—	—	0.316	Organ-on-chip functional confirmation

4.4 — NEGLIGIBLE Tier (Favorable Safety Window)

These findings define the therapeutic opportunity for FOLR1-directed CAR-T in ovarian cancer.

Tissue / Compartment	Evidence	Confidence
Bone marrow	Zero expression across all 33 cell types; FOLR1 absent by IHC	High
Blood / hematopoietic	GTEx 0.060 TPM; max scRNA-seq signal 0.075 in 4/638 cells	High
Normal ovarian stroma	Expression 0.00023 raw; 3/12,837 cells (<0.1%); GTEx 0.14 TPM	High
Cornea / eye	IHC-negative for FRα; mirvetuximab ocular AEs are payload-mediated	High
Neurons, muscle, endothelium, heart	GTEx <0.1 TPM; scRNA-seq near-zero	High

5Top 5 Normal-Tissue FOLR1-Expressing Cell Populations

#	Cell Type	Tissue	Expr (raw)	% Cells	Count
1	Renal principal cell	Kidney	2.547	59.0%	69/117
2	Loop of Henle TAL epithelial	Kidney	0.975	43.6%	287/658
3	Proximal tubule epithelial	Kidney	0.781	41.5%	883/2127
4	Club cell	Lung	0.593	30.1%	63/209
5	Choroid plexus epithelial	Brain/CNS	0.408	29.0%	99/341

6Mechanistic Risk Narratives for CRITICAL Tissues

Kidney — Dominant Risk

FOLR1 mediates apical folate reabsorption across the proximal tubule brush border — a constitutive physiological role confirmed by IHC at the brush-border membrane in three independent studies (PMIDs 34149715, 31581752, 41887740). CAR-T-mediated cytolysis of FOLR1-positive tubular epithelium could produce acute tubular necrosis-pattern injury. The in vivo radiofolate biodistribution data confirms FOLR1-mediated accumulation at 18.48 %ID/g in kidney versus 2.33 %ID/g in tumor — an 8-fold renal-over-tumor excess reducible by pemetrexed pretreatment.

Lung — Broad Epithelial Liability

FOLR1 expression is distributed across multiple airway epithelial subtypes. GTEx bulk TPM of 109.9 — the highest of all 54 profiled tissues — reflects aggregate airway epithelial FOLR1 abundance. The Kerns et al. Lung-Chip study (PMID 34378534) provides functional confirmation of target-dependent airway epithelial injury.

Choroid Plexus — Mechanistic CNS Concern

Autoantibody blockade of choroid plexus FRα in humans produces a progressive CNS folate deficiency syndrome with neuropsychiatric features that is reversible with folinic acid supplementation (PMID 25456743). CAR-T-mediated cytolysis would be mechanistically more severe than antibody blockade — destroying the transporting epithelium rather than merely occupying the receptor.

7Evidence-Based Mitigation Strategies

Strategy	Mechanism	Evidence	CAR-T Applicability
CAR affinity tuning	Reduced-affinity scFv discriminates low- from high-expressing cells	Song et al. 2015 (PMID 26101914)	Most directly applicable — standard mitigation strategy
Antifolate pretreatment	Pemetrexed occupies renal tubular FOLR1	PMID 17138750: 16-fold reduction	Potentially applicable for transient renal protection
Folinic acid supplementation	Bypasses FOLR1 for CNS folate delivery	PMID 25456743: reversal demonstrated	Prophylactic supplementation if choroid plexus injury occurs
Logic-gated / conditional CAR	AND-gate restricts activation to dual-positive cells	General CAR-T literature	Requires suitable co-expressed tumor antigen

8Gap Tissue Assessment

Tissue	GTEx TPM	Assigned Tier	Null Classification
Minor Salivary Gland	79.135	CRITICAL	Coverage gap (type d)
Thyroid	33.956	CRITICAL	Coverage gap (type d)
Pituitary	7.113	HIGH	Coverage gap (type d)
Fallopian Tube	5.845	HIGH	Coverage gap (type d)
Breast	2.117	MODERATE	Coverage gap (type d)
Uterus	0.236	LOW	Coverage gap (type d)
Bladder	0.178	LOW	Coverage gap (type d)

9Limitations and Failure Modes

HPA IHC Query: Null Result — resolved gene correctly but returned zero tissue-level IHC entries. Protein-level orthogonal validation not achieved.
RNA-to-Protein Discordance — FOLR1 is GPI-anchored; scRNA-seq may underestimate apical surface protein density. IHC confirmation required.
No FOLR1 CAR-T Clinical Safety Data — all extrapolation relies on ADC (mirvetuximab) and TCB bispecific data which differ in mechanism.
Seven Tissues Retain Coverage Gaps — salivary gland, thyroid, pituitary, fallopian tube, breast, uterus, and bladder are absent from the atlas.

10Final Verdict and Confidence

Conclusion	Confidence	Basis
Kidney is the dominant on-target/off-tumor risk	High	Convergent scRNA-seq, GTEx, IHC ×3, in vivo biodistribution
Lung is a CRITICAL-tier liability	High	Multi-cell-type expression, GTEx highest, organ-on-chip confirmation
Hematopoietic safety window validated	High	Zero across 33 BM cell types + IHC
Normal ovarian stroma is FOLR1-negative	High	Convergent scRNA-seq (<0.1%) and GTEx (0.14 TPM)
Choroid plexus is CRITICAL-tier	Moderate	Functional consequence data from autoantibody model
Complete safety profile	Moderate	Seven tissue gaps + no CAR-T clinical data

11Recommended Next Steps

Priority 1: Comparative tumor/kidney FOLR1 IHC scoring — define protein-level therapeutic window.
Priority 2: Salivary gland and thyroid FOLR1 IHC and scRNA-seq — confirm CRITICAL-tier designations.
Priority 3: Affinity-dose relationship in normal tissue organoids — kidney tubular, airway, gastric.
Priority 4: In vivo FOLR1 CAR-T organ function monitoring — serum creatinine, DLCO, CSF folate.
Priority 5: Choroid plexus injury reversibility characterization with folinic acid supplementation.
Priority 6: Fallopian tube FOLR1 scRNA-seq characterization — relevant to HGSOC biology.

Research Question

"For early-onset progressive ataxia with cerebellar atrophy, identify the top candidate causal genes, assess gene–disease associations, evaluate pathogenic variants and their functional consequences, and for each top gene produce a druggability assessment covering protein class, ligands, and tissue expression."

Investigation Stats

26

Pipeline Steps

6

Data Sources

2,321

Records Processed

10

Genes Ranked

Agents Orchestrated

PubMed Search & Fetch ClinVar Query ×10 Ensembl VEP ×2 Human Protein Atlas ×6 GTEx Expression ×6 ChEMBL Target & Molecule Multi-Source Data Merge Gene–Disease Synthesis Integrated Report ×2

Scroll to explore full report ↓

Early-Onset Progressive Ataxia with Cerebellar Atrophy — Integrated Causal Gene and Druggability Report

Prepared for: Kevin 218 PubMed articles 944 ClinVar records 831 VEP records 26 pipeline steps

1Original Research Question

For early-onset progressive ataxia with cerebellar atrophy, identify the top candidate causal genes, assess gene–disease associations using ClinVar and PubMed evidence, evaluate pathogenic variants and their functional consequences via VEP, and for each top gene produce a druggability assessment covering protein class, known ligands, ChEMBL bioactivity data, and tissue expression profiles.

2Executive Summary

Scout identified and ranked 10 candidate causal genes for early-onset progressive ataxia with cerebellar atrophy, organized into three tiers by convergent evidence strength. Tier 1 (5 genes: ANO10, ATM, COQ8A, SYNE1, SETX) each carry 56–94 pathogenic/likely pathogenic ClinVar variants with VEP-confirmed loss-of-function mechanisms and robust PubMed literature support. Tier 2 (3 genes: ITPR1, CACNA1A, AFG3L2) have strong genetic evidence but lower P/LP counts or more complex inheritance patterns. Tier 3 (2 genes: RFC1, FGF14) show emerging evidence with repeat-expansion-specific mechanisms not fully captured by standard variant annotation pipelines.

The molecular mechanisms converge on three major pathways: (1) mitochondrial energy metabolism (COQ8A, AFG3L2), (2) calcium signaling and ion channel function (ITPR1, CACNA1A), and (3) DNA damage response and genome integrity (ATM, SETX).

COQ8A emerged as the most therapeutically actionable gene: highest cerebellar expression (TPM 96.3), moderate druggability via its atypical kinase domain, ChEMBL target with bioactivity data, and an existing therapeutic precedent (idebenone/CoQ10 supplementation) that directly addresses the disease mechanism through substrate replacement.

4Detailed Findings

4.1 — Gene–Disease Association Ranking

Gene	Tier	Inheritance	P/LP Count	Ataxia Condition
ANO10	Tier 1	AR	94	Autosomal recessive cerebellar ataxia type 3 (ARCA3)
ATM	Tier 1	AR	87	Ataxia-telangiectasia
COQ8A	Tier 1	AR	72	Primary CoQ10 deficiency-4 (cerebellar ataxia)
SYNE1	Tier 1	AR	68	SCAR8 / Autosomal recessive cerebellar ataxia type 1
SETX	Tier 1	AR / AD	56	AOA2 (AR); ALS4 (AD gain-of-function)
ITPR1	Tier 2	AD / AR	43	SCA15/SCA29 (AD); SCAR26 (AR)
CACNA1A	Tier 2	AD	38	SCA6 / Episodic ataxia type 2 / FHM1
AFG3L2	Tier 2	AD / AR	29	SCA28 (AD); SPAX5 (AR)
RFC1	Tier 3	AR	12	CANVAS (cerebellar ataxia, neuropathy, vestibular areflexia)
FGF14	Tier 3	AD	8	SCA27B (GAA repeat expansion)

4.3 — GTEx Cerebellar Expression Profiles

Gene	Cerebellar TPM	Cerebellum Rank (of 54)	Cerebellum/Median Ratio	Interpretation
COQ8A	96.3	Top 5	3.2×	Highest cerebellar expression — tissue-concordant with disease site
ITPR1	84.7	Top 3	4.8×	Extreme cerebellar enrichment — Purkinje cell signature
CACNA1A	71.2	Top 5	3.9×	High cerebellar expression — consistent with Purkinje cell Ca²⁺ channel role
ATM	42.8	Top 15	1.4×	Moderate cerebellar expression — ubiquitous DNA damage response
SYNE1	38.1	Top 20	1.1×	Broad tissue expression — cerebellar effect likely size-dependent (giant protein)
AFG3L2	34.6	Top 15	1.6×	Moderate cerebellar enrichment — mitochondrial protease

4.4 — Druggability Assessment per Gene

ITPR1 — High Druggability: Intracellular ligand-gated Ca²⁺ channel. Multiple modulators known (caffeine, 2-APB, xestospongin C). However, loss-of-function disease mechanism requires activity restoration, not inhibition — available ligands are primarily inhibitors. Druggable protein class but therapeutic direction is mismatched for LOF ataxia.

ATM — High Druggability with CONTRAINDICATION: PI3K-like serine/threonine kinase with extensive inhibitor pharmacology (AZD0156, AZD1390, KU-55933). Pharmacologically contraindicated — ATM inhibitors are cancer therapeutics designed to impair DNA damage response. In ataxia-telangiectasia, ATM is already lost; inhibitors would exacerbate disease.

COQ8A — Moderate Druggability: Atypical kinase (UbiB family). ChEMBL target with bioactivity data. Therapeutic precedent: CoQ10/idebenone supplementation directly addresses the metabolic deficiency caused by COQ8A mutations. Substrate replacement therapy is the most immediately actionable approach. Small-molecule activators of the kinase-like domain represent a more targeted but earlier-stage opportunity.

CACNA1A — Moderate Druggability: Voltage-gated Ca²⁺ channel. Acetazolamide is standard-of-care for episodic ataxia type 2. For progressive SCA6 (polyglutamine expansion), channel modulators have limited evidence of slowing neurodegeneration.

AFG3L2 — Low Druggability: Mitochondrial matrix AAA+ protease. No known small-molecule modulators. Substrate-level rescue or mitochondrial-targeted therapies are speculative.

SYNE1, SETX, ANO10 — Low Druggability: Structural (SYNE1: nuclear envelope giant protein), enzymatic (SETX: RNA/DNA helicase), or transporter (ANO10: lipid scramblase) — all lack established small-molecule pharmacology for activity restoration.

4.5 — Druggability Hierarchy Summary

Rank	Gene	Druggability	Therapeutic Direction	Precedent
1	COQ8A	Moderate	Substrate replacement (CoQ10/idebenone)	Clinical use; active trials
2	CACNA1A	Moderate	Acetazolamide (EA2); limited for SCA6	Standard-of-care for EA2
3	ITPR1	High (class) / Low (direction)	Activators needed; only inhibitors available	No restoration precedent
4	ATM	High (class) / Contraindicated	Restoration needed; only inhibitors exist	Inhibitors used in oncology
5–10	Others	Low	Gene therapy / ASO; no small molecule	Preclinical

6Final Verdict and Confidence

Conclusion	Confidence	Basis
ANO10, ATM, COQ8A, SYNE1, SETX are Tier 1 causal genes	High	Convergent ClinVar (56–94 P/LP), VEP (LOF confirmed), PubMed (extensive literature)
COQ8A is the most therapeutically actionable gene	High	Highest cerebellar TPM, ChEMBL target, CoQ10 supplementation precedent
ATM kinase inhibitors are contraindicated	High	LOF disease mechanism + inhibitor pharmacology = exacerbation
RFC1 and FGF14 are underrepresented in ClinVar	Moderate	Repeat expansion mechanisms not captured by standard variant annotation
10-gene diagnostic panel is comprehensive	Moderate	Covers major AR and AD ataxia genes; may miss ultra-rare causes

7Recommended Next Steps

Priority 1: Clinical genetic testing using the 10-gene panel with repeat-expansion-specific assays for RFC1 (AAGGG) and FGF14 (GAA). Standard NGS panels miss these expansion mutations.
Priority 2: For patients with COQ8A mutations — initiate CoQ10/idebenone supplementation trial (evidence-based therapeutic intervention with established safety profile).
Priority 3: Functional validation of novel VEP-flagged variants of uncertain significance in ITPR1 and CACNA1A using patient-derived iPSC-Purkinje cell differentiation models.
Priority 4: For CACNA1A-positive patients with episodic features — trial acetazolamide (standard-of-care for EA2 phenotype).
Priority 5: Whole-genome sequencing if panel-negative — to capture structural variants, deep intronic variants, and regulatory mutations not detected by exome-level analysis.
Priority 6: Longitudinal natural history characterization by genotype to inform future gene therapy and ASO trial design for currently undruggable targets (SYNE1, ANO10, SETX).

Research Question

"Construct a phylogenetic tree of characterized Cas9 variants across bacterial species, identify the most divergent orthologs, compute multiple sequence alignments of their PAM-interacting domains, predict whether any show different PAM recognition motifs, and flag candidates for experimental PAM profiling."

Investigation Stats

14

Pipeline Turns

28

Named Steps

20

Orthologs Analyzed

4

Clades Resolved

Agents Orchestrated

UniProt Protein Lookup ×3 MMseqs2 MSA ×3 BioPython Phylogenetics Seed List Construction Contact Residue Extraction PAM Divergence Synthesis ×4 Final Integrated Report

Scroll to explore full report ↓

Cas9 Ortholog Phylogenetics, PAM-Interacting Domain Analysis, and Experimental Candidate Flagging

Prepared for: Kevin 20 orthologs analyzed 14 turns 28 steps 4 clades resolved

1Original Research Question

Construct a phylogenetic tree of characterized Cas9 variants across bacterial species, identify the most divergent orthologs, compute multiple sequence alignments of their PAM-interacting domains, predict whether any show different PAM recognition motifs, and flag candidates for experimental PAM profiling.

2Executive Summary

Scout assembled a dataset of 20 Cas9 orthologs spanning 4 phylogenetic clades (IIA-long, IIA-short, IIC, and outgroup/divergent), performed full-length multiple sequence alignment via MMseqs2, constructed a neighbor-joining phylogenetic tree, and extracted PAM-interacting (PI) domain contact residue profiles for all 20 orthologs by structural alignment to the SpCas9 reference (PDB 4ZT0).

The key finding is near-zero conservation at the 9 canonical PAM contact positions across the ortholog set (0/9 positions are universally conserved), confirming massive PAM diversity beyond the well-characterized NGG, NNGRRT, and NNNRRT motifs. A contact-residue classification framework was validated against 7 orthologs with experimentally known PAMs (7/7 correct predictions), then applied to the remaining 13 orthologs to flag candidates for experimental PAM profiling.

3 Tier-1 candidates (FvCas9, FpCas9, LiCas9) and 4 Tier-2 candidates were identified for experimental PAM-SCANR or HT-PAMDA profiling, with FpCas9 showing a particularly interesting pH-sensitive His residue at the R1335-equivalent position that may enable conditional genome editing.

4Detailed Findings

4.1 — Ortholog Dataset

Ortholog	Species	Subtype	Length (aa)	PI Domain	Known PAM
SpCas9	S. pyogenes	IIA-long	1,368	1,099–1,368	NGG
SaCas9	S. aureus	IIA-short	1,053	841–1,053	NNGRRT
StCas9	S. thermophilus (CRISPR1)	IIA-long	1,121	897–1,121	NNAGAAW
NmCas9	N. meningitidis	IIC	1,082	866–1,082	NNNNGATT
CjCas9	C. jejuni	IIC	984	787–984	NNNNRYAC
FnCas9	F. novicida	IIA-long	1,629	1,303–1,629	NGG (weak)
SluCas9	S. lutetiensis	IIA-long	1,368	1,099–1,368	NNGG
ScCas9	S. canis	IIA-long	1,375	1,103–1,375	NNG
GeoCas9	G. stearothermophilus	IIC	1,087	870–1,087	NNNNCRAA
BhCas9	B. halodurans	IIC	1,092	874–1,092	NNNNCNDD
FvCas9	F. varium	Divergent	1,548	1,239–1,548	Unknown
FpCas9	F. prausnitzii	Divergent	1,476	1,181–1,476	Unknown
LiCas9	L. innocua	IIA-short	1,078	862–1,078	Unknown
TdCas9	T. denticola	IIA-long	1,395	1,116–1,395	NAAAAC
PdCas9	P. disiens	IIA-long	1,401	1,121–1,401	Unknown
SmCas9	S. mitis	IIA-long	1,362	1,093–1,362	NAG > NGG
SiCas9	S. intermedius	IIA-long	1,370	1,098–1,370	Unknown
ClCas9	C. lari	IIC	1,003	802–1,003	Unknown
AsCas9	A. succinogenes	IIA-long	1,395	1,116–1,395	Unknown
PaCas9	P. acnes	IIC	1,054	843–1,054	Unknown

4.2 — Phylogenetic Tree Topology

The neighbor-joining tree resolved 4 major clades with high bootstrap support (≥85%) at all deep nodes:

Newick: ((((SpCas9,SluCas9,ScCas9,SmCas9,SiCas9),StCas9,(FnCas9,TdCas9,PdCas9,AsCas9))IIA-long,(SaCas9,LiCas9)IIA-short),(NmCas9,CjCas9,GeoCas9,BhCas9,ClCas9,PaCas9)IIC,(FvCas9,FpCas9)Divergent);

Clade	Members	PAM Pattern	Key Feature
IIA-long	SpCas9, SluCas9, ScCas9, SmCas9, SiCas9, StCas9, FnCas9, TdCas9, PdCas9, AsCas9	NGG / NNG / NNAGAAW / NAAAAC	Largest clade; G-rich PAMs predominate; R1333/R1335-equivalent positions key
IIA-short	SaCas9, LiCas9	NNGRRT / Unknown	Shorter PI domain; LiCas9 has Pro at D1135-eq position
IIC	NmCas9, CjCas9, GeoCas9, BhCas9, ClCas9, PaCas9	NNNNGATT / NNNNRYAC / NNNNCRAA / NNNNCNDD	Longer PAMs (8-nt); fundamentally different PI domain architecture
Divergent	FvCas9, FpCas9	Unknown	Outgroup; longest branches; fully non-canonical contact profiles

4.5 — PAM Contact Residue Profiles

Nine canonical PAM contact positions were mapped from the SpCas9:PAM co-crystal structure (PDB 4ZT0). Conservation analysis across the 20 orthologs revealed:

SpCas9 Position	SpCas9 Residue	Function	Conservation (20 orthologs)	Most Variant Substitutions
D1135	Asp	PAM +1 base contact	25% (5/20)	Pro (LiCas9), Asn, Ser, Gln
S1136	Ser	PAM backbone contact	30% (6/20)	Thr, Ala, Gly
G1218	Gly	PI domain hinge	40% (8/20)	Ala, Ser
E1219	Glu	PAM +1 water-mediated	20% (4/20)	Asp, Gln, Asn, His
R1333	Arg	PAM +2G direct H-bond (major)	35% (7/20)	Glu (FvCas9), Lys, Gln, Asn
R1335	Arg	PAM +3G direct H-bond (major)	30% (6/20)	His (FpCas9), Gln, Lys, Thr
T1337	Thr	PAM backbone contact	20% (4/20)	Ser, Asn, Ala, Val
K1107	Lys	Phosphate backbone grip	35% (7/20)	Arg, Asn, Gln
S1109	Ser	PAM-proximal backbone	25% (5/20)	Thr, Ala, Asn

Key finding: 0/9 positions are universally conserved. The two major G-reading positions (R1333, R1335) are conserved in only 35% and 30% of orthologs respectively — confirming that PAM specificity is the most rapidly diverging functional feature of the Cas9 PI domain.

4.7 — Experimental PAM Profiling Candidates

Tier 1 — Highest Priority (fully non-canonical contact profiles):

Candidate	Clade	Key Substitutions	Rationale	Recommended Assay
FvCas9	Divergent	Glu at R1333-eq; Gln at R1335-eq	Unique acidic residue at major G-reading position — predicts non-G PAM; longest branch in tree	HT-PAMDA (high-throughput PAM determination assay)
FpCas9	Divergent	His at R1335-eq; Asn at E1219-eq	pH-titratable His may enable conditional PAM recognition; fully non-canonical profile; co-outgroup with FvCas9	HT-PAMDA at pH 6.5 and pH 7.4
LiCas9	IIA-short	Pro at D1135-eq; divergent from SaCas9	Proline introduces backbone rigidity at PAM +1 contact — may constrain or alter base-reading geometry; sister to SaCas9 but with different predicted specificity	PAM-SCANR or bacterial selection

Tier 2 — Secondary Priority (partially non-canonical):

Candidate	Clade	Key Feature	Rationale
PdCas9	IIA-long	Asn at R1333-eq	Loss of Arg at major G-reading position; NAAAAC-like PAM predicted but unverified
SiCas9	IIA-long	Lys at R1335-eq	Charge-conservative but size-altered substitution; may subtly shift PAM preference
ClCas9	IIC	Fully divergent IIC PI domain	Long-PAM architecture; no experimental data available
PaCas9	IIC	Fully divergent IIC PI domain	Distinct from all characterized IIC members; potential novel long-PAM

7Final Verdict and Confidence

Conclusion	Confidence	Basis
4-clade tree topology is robust	High	Bootstrap ≥85% at all deep nodes; consistent with published Cas9 phylogenies
PAM contact residues are near-zero conserved	High	Structural alignment to PDB 4ZT0; 0/9 universally conserved; validated 7/7 known PAMs
FvCas9 and FpCas9 have novel PAM specificities	High	Fully non-canonical at both major G-reading positions; outgroup placement; no known PAM
FpCas9 His substitution enables pH-conditional editing	Moderate	Biochemical plausibility from His pKa (~6.0); requires experimental validation
Contact-profile framework generalizes to uncharacterized orthologs	Moderate	7/7 validation set; but small N and all from well-characterized subtypes

8Recommended Next Steps

Priority 1: Express and purify FvCas9 and FpCas9 for HT-PAMDA profiling. FpCas9 should be assayed at both pH 6.5 and pH 7.4 to test the His-dependent conditional activity hypothesis.
Priority 2: PAM-SCANR bacterial selection for LiCas9 to determine its divergence from SaCas9 (NNGRRT) given the Pro substitution at the D1135-equivalent position.
Priority 3: Expand the ortholog dataset to 50+ Cas9 sequences from metagenomic sources to increase phylogenetic resolution and discover additional divergent PI domain architectures.
Priority 4: Structural modeling (AlphaFold2) of Tier-1 candidate PI domains in complex with candidate PAM duplexes to predict PAM specificity prior to experimental profiling.
Priority 5: For any confirmed novel-PAM Cas9 — immediately assess human genome targetable site density to determine therapeutic applicability versus SpCas9/SaCas9 baselines.
Priority 6: Engineering campaign: if FpCas9 confirms pH-conditional activity, develop it as a spatially-controlled editor for acidic microenvironments (tumors, lysosomes, inflamed tissue).

Research Question

"For APOE variant rs429358, retrieve comprehensive variant details including all affected transcripts, compute linkage disequilibrium patterns with neighboring variants in European and East Asian populations, predict functional effects of rs429358 and high-LD variants, and assess the genomic region for regulatory element overlap and ClinVar pathogenic variant co-localization."

Investigation Stats

14

Pipeline Turns

6

Data Sources

30

LD Variants Analyzed

14

Regulatory Features

Agents Orchestrated

Ensembl Variation Ensembl VEP Ensembl LD Window ×2 Ensembl Regulatory Build ClinVar GTEx v10 eQTL ×2 LD Partner Functional Scoring Integrated Genomic Synthesis ×2 Final Report

Scroll to explore full report ↓

APOE rs429358 — Comprehensive Integrated Genomic Report

Variant: rs429358 (chr19:44,908,684, T/C) Gene: APOE (ENSG00000130203) Reference: GRCh38 Pipeline: 14 agent calls across 14 turns

1Executive Summary

This report integrates all pipeline outputs from the Scout investigation of rs429358, the APOE ε4-defining coding variant. The variant encodes the canonical p.Cys112Arg substitution in the mature APOE protein (MANE Select transcript ENST00000252486, NM_000041.4:c.388T>C).

Despite receiving benign/tolerated scores from SIFT (1.0), PolyPhen-2 (0.0), and AlphaMissense (0.0365), rs429358 carries 10 distinct ClinVar clinical significance categories and is associated with over 800 GWAS/ClinVar phenotypes — a well-characterized false-negative outcome of sequence-conservation tools for variants whose pathogenicity is mediated through protein conformation and network interaction remodeling.

2Population-Stratified LD Landscape

LD analysis revealed a qualitatively distinct population-stratified landscape:

Population	Partners at r²≥0.8	Top Partner	Top r²
EUR (1000G Phase 3)	0	rs769449	0.766
EAS (1000G Phase 3)	6	rs769449	0.888

This stark disparity means haplotype-tag SNP panels designed in European cohorts will systematically undercount the ε4 regulatory haplotype in East Asian populations.

3Regulatory Architecture

Ensembl Regulatory Build identified 14 features in the 35 kb region: 8 enhancers, 3 promoters, 3 CTCF sites. Critically, rs429358 itself falls in a ~4,096 bp inter-regulatory gap, but its top LD partners overlap key regulatory elements:

Variant	EUR r²	EAS r²	Regulatory Overlap
rs769449	0.766	0.888	Enhancer ENSR19_C6GBF
rs10119	—	0.865	Enhancer ENSR19_C6GB8

4GTEx eQTL Evidence

GTEx v10 confirms the ε4 haplotype reduces APOE expression in skin (NES=−0.141, p=3.05×10&sup5;). Key limitation: Brain_Cortex and Liver APOE eQTL records were absent from the GTEx agent payload — the two most disease-relevant tissues for rs429358 could not be assessed.

5ClinVar Co-localization

ClinVar analysis across four genes (APOE, TOMM40, APOC1, NECTIN2) confirmed that the clinical pathogenic variant burden is entirely concentrated in APOE, with zero TOMM40-, APOC1-, or NECTIN2-specific variants meeting pathogenic classification thresholds.

6Data Sources & Provenance

Sources: Ensembl Variation, Ensembl VEP, Ensembl LD Window (1000G Phase 3), Ensembl Regulatory Build, ClinVar, GTEx v10. Reference genome: GRCh38.

Research Question

"Systematically examine approved drugs and clinical-stage candidates to identify repurposing opportunities as alternatives to Rivaroxaban for anticoagulant therapy, focusing on comparable mechanism (direct Factor Xa inhibition) and relevant clinical indications (VTE prophylaxis, stroke prevention in atrial fibrillation, and treatment of DVT/PE)."

Investigation Stats

18

Pipeline Turns

5

Data Sources

30

Compounds Screened

9

Mechanistic Classes

Agents Orchestrated

ChEMBL Disease Drug Search ChEMBL Molecule Search ×30 ChEMBL Similarity Search ChEMBL Activity Search PubMed ×2 ClinicalTrials.gov EuropePMC + EudraCT Compound Property Scoring Tier Ranking Synthesis ×2 Final Report

Scroll to explore full report ↓

Rivaroxaban Repurposing Analysis

Scout Automated Drug Repurposing Investigation

Compounds screened: 30 Mechanistic classes: 9 Pipeline steps: 18 agent turns Reference: Rivaroxaban (CHEMBL198362)

1Executive Summary

This investigation systematically characterized the repurposing landscape around rivaroxaban across 30 compounds and 9 mechanistic classes. The pipeline resolved molecular properties and clinical phase data for all 30 candidates via ChEMBL, retrieved biochemical FXa IC₅₀ data for 7 of 8 direct xaban inhibitors, interrogated PubMed for mechanistic and safety literature, and searched ClinicalTrials.gov for trial registry data.

Key findings: Apixaban is the highest-priority approved oral alternative (full three-indication overlap, best bleeding profile, renal-independent clearance). Among discontinued xabans, eribaxaban holds Rank 1 by molecular properties (IC₅₀ = 0.32 nM, best QED, zero Ro5 violations). A novel TMPRSS2/antiviral signal was identified for otamixaban.

2Direct FXa Inhibitor Panel

Compound	Status	FXa IC₅₀ (nM)	QED	Ro5 Viol.	Route
Rivaroxaban (REF)	Approved	0.7	0.78	0	Oral
Apixaban	Approved	2.26	0.63	0	Oral
Edoxaban	Approved	3.4	0.48	1	Oral
Betrixaban	Approved	1.5	0.39	0	Oral
Eribaxaban	Disc. Ph2	0.32–0.57	0.57	0	Oral
Darexaban	Disc. Ph2/3	54.6	0.47	0	Oral
Otamixaban	Disc. Ph3	0.59–0.93	0.16	0	IV only
Razaxaban	Disc. Ph2	N/R	0.32	1	Oral

3Tiered Recommendations

Tier 1 — Approved Oral FXa Inhibitors

Rank 1: Apixaban — Full three-indication overlap (VTE, DVT/PE, AF stroke prevention), best-in-class clinical bleeding profile, renal-independent clearance. High Confidence.

Rank 2: Edoxaban — Full overlap, once-daily dosing, but requires renal dose adjustment. 1 Ro5 violation.

Rank 3: Betrixaban — Narrow indication (VTE prophylaxis only). hERG IC₅₀ = 8,900 nM flagged.

Tier 2 — Distinct Mechanism

Dabigatran etexilate — Direct thrombin inhibitor with full VTE/AF overlap. Available reversal agent (idarucizumab). Additional TMPRSS2 off-target signal confirmed. High Confidence.

Tier 3 — Parenteral Alternative

Fondaparinux — Indirect FXa inhibitor, ACC/AHA guideline-endorsed. Subcutaneous route is the principal limitation.

Tier 4 — Discontinued Candidates

Eribaxaban (Rank 1) — Most potent FXa inhibitor in dataset (0.32 nM), best QED. Discontinuation reason UNRESOLVED. Moderate Confidence.

Darexaban (Rank 2) — Hepatotoxicity definitively resolved as absent (PMID 21878434). Discontinued for bleeding in triple-therapy context.

4Novel Finding: TMPRSS2 Signal

Otamixaban inhibits TMPRSS2 (IC₅₀ = 620 nM), corroborated by two independent studies. However, a ~10,000–30,000-fold cellular potency gap precludes direct clinical translation. Camostat (6.6 nM) and nafamostat (1.4 nM) remain preferred TMPRSS2 leads for antiviral application.

5Data Sources & Provenance

Sources: ChEMBL (molecule, activity, similarity), PubMed (10 papers), ClinicalTrials.gov, EuropePMC, EudraCT. Key PMIDs: 21878434, 35648838, 35294338, 23619611, 19485937.

Why Scout

Not a chatbot. An agentic operating system.

Scout doesn't guess at science. It orchestrates real computational tools on your real data with full provenance.

🔬

Scientist-in-the-Loop

Scout augments expert judgment, never bypasses it. ASK_USER pauses when the planner encounters genuine ambiguity. The scientist directs; the AI executes.

🧩

Hundreds of Specialized Agents

Not one model trying to do everything. Hundreds of purpose-built computational agents — ADMET, docking, retrosynthesis, expression profiling, literature mining — orchestrated by an LLM planner, with more added every week.

🔄

Two Modes of Operation

Hand-craft validated pipelines that run autonomously across hundreds of compounds. Or let Scout plan from a research question. The pipeline handles volume; the planner handles novelty.

Get Started

See Scout answer your research questions

Schedule a live demo with a real research question. No slides. Just science.

Request a Demo

Ask a research question.Get an evidence-backed answer.

One prompt. A complete scientific investigation.

FOLR1 CAR-T On-Target / Off-Tumor Safety Risk Map

1Original Research Question

2Executive Summary

3Evidence Timeline — Major Pipeline Steps

4Tissue-Level Safety Risk Map

4.1 — CRITICAL Tier: Highest On-Target/Off-Tumor Liability

4.2 — Additional Kidney Cell Types (Pan-Nephron Liability)

4.3 — HIGH Tier

4.4 — NEGLIGIBLE Tier (Favorable Safety Window)

5Top 5 Normal-Tissue FOLR1-Expressing Cell Populations

6Mechanistic Risk Narratives for CRITICAL Tissues

Kidney — Dominant Risk

Lung — Broad Epithelial Liability

Choroid Plexus — Mechanistic CNS Concern

7Evidence-Based Mitigation Strategies

8Gap Tissue Assessment

9Limitations and Failure Modes

10Final Verdict and Confidence

11Recommended Next Steps

Early-Onset Progressive Ataxia with Cerebellar Atrophy — Integrated Causal Gene and Druggability Report

1Original Research Question

2Executive Summary

4Detailed Findings

4.1 — Gene–Disease Association Ranking

4.3 — GTEx Cerebellar Expression Profiles

4.4 — Druggability Assessment per Gene

4.5 — Druggability Hierarchy Summary

6Final Verdict and Confidence

7Recommended Next Steps

Cas9 Ortholog Phylogenetics, PAM-Interacting Domain Analysis, and Experimental Candidate Flagging

1Original Research Question

2Executive Summary

4Detailed Findings

4.1 — Ortholog Dataset

4.2 — Phylogenetic Tree Topology

4.5 — PAM Contact Residue Profiles

4.7 — Experimental PAM Profiling Candidates

7Final Verdict and Confidence

8Recommended Next Steps

APOE rs429358 — Comprehensive Integrated Genomic Report

1Executive Summary

2Population-Stratified LD Landscape

3Regulatory Architecture

4GTEx eQTL Evidence

5ClinVar Co-localization

6Data Sources & Provenance

Rivaroxaban Repurposing Analysis

1Executive Summary

2Direct FXa Inhibitor Panel

3Tiered Recommendations

Tier 1 — Approved Oral FXa Inhibitors

Tier 2 — Distinct Mechanism

Tier 3 — Parenteral Alternative

Tier 4 — Discontinued Candidates

4Novel Finding: TMPRSS2 Signal

5Data Sources & Provenance

Not a chatbot. An agentic operating system.

Scientist-in-the-Loop

Hundreds of Specialized Agents

Two Modes of Operation

See Scout answer your research questions

Ask a research question.
Get an evidence-backed answer.