project-idea

BF-BOLD Trial Simulator: Interactive Phase I Dose-Finding Tool

The BF-BOLD design improves safety and activity assessment in Phase I trials without complex statistical modeling — but adoption requires trial teams to understand and simulate the design before committing to it.

Build an open-source, browser-based simulator for the BF-BOLD (and comparison BOLD/3+3) Phase I trial designs. Users would input trial parameters — number of dose levels, target toxicity rate, activity thresholds, cohort sizes — and the simulator would run Monte Carlo simulations showing the expected distribution of outcomes: recommended Phase II dose selection accuracy, overdose rates, and trial duration.

The tool would be aimed at biostatisticians and clinical trialists who want to compare design options before protocol submission. Side-by-side simulation of BF-BOLD vs. conventional 3+3 vs. BOIN vs. mTPI would help teams make evidence-based design decisions and produce output directly useful for protocol justification sections.

A secondary output would be a plain-language trial design summary — useful for IRB submissions and grant applications where committee members may not be statisticians. This kind of simulation tool significantly lowers the barrier to adopting novel trial designs, which is currently the biggest obstacle to real-world BF-BOLD implementation.

Who Is This For?

Biostatisticians, clinical trial designers, oncology research coordinators, and pharma/biotech early development teams.

Skills & Tools Needed

R or Python for Monte Carlo simulation of dose-escalation designs
Familiarity with Phase I trial design literature (BOLD, BOIN, mTPI, 3+3)
Shiny (R) or Streamlit (Python) for interactive web interface
Statistical visualization (ggplot2, Plotly)
Technical writing for plain-language protocol summary generation

Feasibility

high — Monte Carlo simulation of dose-escalation designs is well-understood and R packages already exist for related designs; adapting them for BF-BOLD is a defined, bounded software task.

Inspired by: Backfill Bayesian Ordered Lattice Design for Phase I Clinical Trials

Landmark DCIS trial, CRC microbiome & AML combos

Two themes cut across this week’s 236 posts. The first is de-escalation: researchers are building rigorous evidence that patients can safely do less — skip lymph node surgery, forgo radiotherapy, watch pre-invasive lesions instead of cutting them out. The landmark result here is a randomized trial confirming that active surveillance

Colorectal Cancer Microbiome Discovery Platform

Analysis of 9,000+ patients revealed that only colorectal cancer — not other cancer types — consistently harbors distinct microbial communities, suggesting the gut microbiome has a unique relationship with colorectal carcinogenesis that could be exploited for diagnostics and treatment.

GBM Tumor Microenvironment Drug Target Atlas

Non-cancerous brain support cells were found to promote glioblastoma growth through paracrine signaling, and blocking this communication dramatically slowed tumor growth — revealing that GBM's microenvironment contains targetable vulnerabilities that are being systematically overlooked by drug…

BET Inhibitor Target Selectivity Database

BRD2 and BRD4 have fundamentally distinct functions — BRD2 prepares genes for activation while BRD4 triggers transcription — meaning that pan-BET inhibitors have been disrupting gene regulation in counterproductive ways, and the field now needs selective BRD4 inhibitors.