Skip to content

Wiki_Build_Output.md

Latest commit

 

History

History
101 lines (74 loc) · 7.03 KB

Wiki_Build_Output.md

File metadata and controls

101 lines (74 loc) · 7.03 KB

BSGOU Project Report & Workspace Documentation

This document compiles the structured knowledge base from the BSGOU LLM-Wiki, representing a complete overview of the Bioinformatics Study Group Okayama University (BSGOU) workspace, codebase, and scientific methodologies.

1. Executive Summary & BSGOU Overview

BSGOU (Bioinformatics Study Group Okayama University) is an international community of students, researchers, clinicians, engineers, and industry professionals. The organization is based at Okayama University and focuses on bridging biological data with theoretical models of living systems.

Mission

BSGOU's mission is to collaboratively tackle complex biological problems, develop open-source tools, and train the next generation of bioinformaticians. It promotes open, reproducible science to turn big data into meaningful, actionable insights.

Strategic Goals

  1. Foster Interdisciplinary Collaboration: Uniting clinical knowledge with engineering optimization.
  2. Embrace Systems Biology: Building multi-scale models of complex biological systems.
  3. Upskill Members: Providing hands-on tutorials for both computational and wet-lab techniques.
  4. Outreach & Leadership: Contributing to global standards and ethics in bioinformatics.

2. Core Values

BSGOU operates under 9 core values, which guide all community, research, and coding contributions:

  • Serve the People: Prioritizing community needs and public health impact.
  • Theory-Driven Understanding: Championing mathematical frameworks over simple data accumulation.
  • Transdisciplinary Collaboration: Breaking silos between wet-lab biology and engineering.
  • Education for Empowerment: Upskilling members at all levels.
  • Scientific Integrity: Ensuring strict verification, quality control, and reproducible pipelines.
  • Global Inclusivity: Maintaining a diverse international network.
  • Open Innovation: Promoting open source, open data (FAIR principles), and shared repositories.
  • Visionary Leadership: Preparing the next generation of bioinformatics pioneers.
  • Everyone Can Contribute: Supporting a culture where all corrections, comments, and pull requests are verified, tracked, and rewarded.

3. Reconstructed Historical Timeline

Below is the verified timeline of milestones for the BSGOU codebase and public blog releases:

  • 2025-05-30 [verified]: Launch of the BSGOU public website and publishing of the "About Our Logo" post.
  • 2025-06-01 [verified]: Publication of the GlycoRNA cell-surface imaging journal club summaries across English, Japanese, and Chinese translations.
  • 2025-06-08 [verified]: Stereo-seq CID coordinates to ATGC barcode mapping guide published ("From CID to ATGC").
  • 2025-06-10 [verified]: Publication of the Zebrafish genomic history article detailing duplication events.
  • 2025-06-12 [verified]: Launch of the Contribution Score Poisson normalization method.
  • 2025-06-15 [verified]: Mathematical proof of qPCR relative quantitation normalization published, accompanied by Excel and GraphPad Prism templates.
  • 2026-06-08 [verified]: Repository onboarded to AROS governance, initializing re_gent and seeding the LLM-Wiki.

4. Codebase Components and Workflows

4.1 Member Directory Scraper

The directory ranking pipeline consists of a cron workflow and a Python parsing script:

  • GitHub Workflow (.github/workflows/update_members.yml): Runs daily at 3 AM UTC. Automatically executes the scraper script, commits the updated HTML roster, and uploads build artifacts.
  • Scraper Script (scripts/fetch_bsgou_members.py):
    • Searches the GitHub API for users with files containing the BSGOU member verification tag.
    • Pulls member contribution counts (Commits, Pull Requests, Issues, Repos) inside the LabOnoM organization.
    • Standardizes contributions to a Contribution Score using Poisson distribution normalization: $$\text{Raw Score} = 5 \times \text{PRs} + 3 \times \text{Issues} + 2 \times \text{Commits} + \text{Repos}$$ $$\text{Final Score} = 0.5 \times \left(\frac{\text{Raw}}{\text{Max Raw}} \times 100\right) + 0.5 \times \left(100 - \text{Poisson.sf}(\text{Raw}-1, \mu) \times 100\right)$$
    • Renders output to members.html and saves raw tables to members.xlsx.

4.2 Utility Tools

  • tools/dir-tree.sh: Directory tree printer excluding node and site packages.
  • tools/diff.sh: Configurations comparator across development and production environments.
  • tools/assert-url.js: Node.js script that dynamically updates relative image pathways in markdown files to point to raw GitHub URLs, preventing rendering errors.

5. Scientific Analyses & Cases

5.1 Visium HD Nuclei Segmentation (GSE284271)

The data folder GSE284271/ documents spatial transcriptomics segmentation workflows:

  • Segmentation: Processes high-resolution H&E stained intestinal tissue using StarDist2D neural network models to segment cell nuclei.
  • Binned Barcode Sorting: Custom python algorithms map sub-micron Visium HD bins to their nearest nuclei centroids.
  • UMI Summation: Gene expression matrices are summed within cell boundaries and exported as Seurat/AnnData HDF5 formats for clustering using Scanpy.

5.2 qPCR Livak Normalization Proof

We mathematically verify that relative fold-changes under Kenneth Livak's $2^{-\Delta\Delta C_T}$ method are scale-invariant.

Performing secondary normalization (dividing individual fold-changes by the arithmetic mean of the control group's $2^{-\Delta\Delta C_T}$) yields identical relative fold-changes as dividing raw $2^{-\Delta C_T}$ expression by the control group's mean:

$$ \frac{2^{-\Delta C_{T, X_i}}}{\frac{1}{n} \sum 2^{-\Delta C_{T, C_i}}} = \frac{2^{-\Delta\Delta C_{T, X_i}}}{\frac{1}{n} \sum 2^{-\Delta\Delta C_{T, C_i}}} $$

As a result, statistical parameters ($P$-values, $t$-statistics, and ANOVA $F$-ratios) remain exactly the same, which is verified using the GraphPad Prism sheets (AM_Calibrator_2DDCt.prism / GM_Calibrator_2DDCt.prism).

5.3 Stereo-seq CID Barcode Translation

  • Barcoding Issue: AI models often suggest simple base-4 modular math conversions for Stereo-seq coordinates, which fail against BAM file sequence reads.
  • Resolution: Actual sequence mapping requires loading MGI's saw mask coordinates (A02598A4.barcodeToPos.h5) using ST_BarcodeMap tool (action 3).
  • Compilation Lesson: Compiling ST_BarcodeMap requires Boost exactly at version 1.73.0. The workspace makefile was adjusted to link local conda include/lib directories correctly: 
    INCLUDE_DIRS = -I$(DIR_INC) -I$(CONDA_PREFIX)/include -I/usr/include/hdf5/serial
LIBRARY_DIRS = -L$(CONDA_PREFIX)/lib -L/usr/lib/x86_64-linux-gnu/hdf5/serial

6. System Engineering and Lessons Learned

  • Jekyll Setup: Ruby and bundle environments are managed in codex.yaml for clean static site previews.
  • Git Push Bypass: The IDE wrapper dummy token is bypassed during pushes by running env -u GITHUB_TOKEN git push origin main to let native system authentication take over.