Gonzales — NF Data Mining & Drug Discovery

Neurofibromatosis data mining collaboration: the first multi-product composition for external science, with CTF NDU grant alignment.

Status: Engaged — product composition mapped, CTF NDU aligned Domain: Drug discovery transcriptomics, neurofibromatosis data mining, One Health cross-species Products: helixVision + healthSpring + initioChem + coralForge (multi-product composition) Springs Fed: wetSpring, healthSpring, neuralSpring, hotSpring Funding: CTF NF Data Utilization Award (NDU) — up to $125K total ($50K Y1 + $75K Y2)


Two Active Threads

Thread 0: GPS Platform Rebuild — tideGlass

The collaborator assigned the GPS platform (Bin Chen Lab) for rebuild as a pseudoSpore — the first concrete deliverable.

Platform: tideGlass (cross-spring biological current parser) Source: Deep learning-based screening and design of novel therapeutics that reverse disease-associated transcriptional phenotype — Cell (2026)

What GPS does: Predicts drug-induced gene expression changes from chemical structures (SMILES to expression profile), then screens for compounds that reverse disease transcriptomic signatures. Three modules:

  • RCL: Robust Collaborative Learning — cleans noisy drug-induced profiles
  • GPS4Drug: Structure to expression prediction + virtual screening
  • MolSearch: Multi-objective lead optimization via MCTS

pseudoSpore target: Full lineage trace, reproduce every figure, rebuild sovereign, package with automated validation checks.

Thread 1: NF Data Mining Project

Neurofibromatosis data mining tied to the Children’s Tumor Foundation. This is the first multi-product composition for an external PI.

The Biological Connection: NF1 loss = hyperactive RAS/MAPK. The collaborator studies JAK/STAT via oclacitinib (782/782 validated checks). NF1 tumors have hyperactivated STAT3. Same signaling biology, different disease context.

Full case study: NF Case Study


Product Composition

ProductRoleExisting Validation
helixVisionGene expression from NF Data PortalwetSpring 6,656+ checks
healthSpringMATRIX drug repurposing (MEK, JAK, mTOR inhibitors)healthSpring 233/233
initioChemMEK inhibitor conformational dynamicshotSpring 190/190
coralForgeNF1 variant structural impactneuralSpring 154 checks

Data Sources

SystemPurposeStatus
NF Data Portal / SynapseGene expression, genomic variants, drug screeningSurvey
Pluto.bio NF datasets~108M data points across 12 datasetsSurvey
PubChem / ChEMBLCompound activity for RAS/MAPK, JAK/STAT targetsSurvey

New Spring Validation Targets

NF work will create validation targets the ecosystem has never seen:

ResultSpring FedValidation Target
NF gene expression signatureswetSpringDifferential expression in NF1 tumors
RAS/MAPK pathway enrichmentwetSpringPathway analysis on NF Data Portal data
NF drug repurposing MATRIX scoreshealthSpringMEK/JAK/mTOR inhibitor ranking
Neurofibromin structural variantsneuralSpringNF1 mutation structural impact
MEK inhibitor binding dynamicshotSpringSelumetinib/trametinib FEL landscapes

The gen5 Pattern

Collaborator brings NF biological question + institutional access + PI authority
    -> Products compose for her domain
    -> Data ingested from NF Data Portal, PubChem, ChEMBL
    -> Computational preliminary data produced (pseudoSpore for NF)
    -> Collaborator uses results for CTF NDU application
    -> If funded: Year 1 bioinformatics -> Year 2 wet lab validation
    -> Results feed back as new spring validation targets
    -> Ecosystem strengthens from external demand

The collaborator did not apply to the ecosystem. She invited it to her science. The NF project is her research question answered by tools that happen to be sovereign, validated, and open. If the ecosystem cannot serve her biology, the architecture is wrong. If it can, gen5 is real.