This topic is best reviewed by a cross-functional panel of Senior Bioinformatics Solutions Architects (to validate the technical integrity of the Slide-tags pipeline) and Executive Talent Strategists in Biopharma (to evaluate the efficacy of the AI-driven career intelligence system). This group possesses the domain expertise to bridge the gap between high-resolution spatial genomics and the strategic labor market within the "Big Pharma" ecosystem.
Abstract:
This documentation details the plops/slide-tag repository, a dual-purpose infrastructure integrating a high-resolution spatial transcriptomics pipeline with an AI-powered career development system. The scientific component processes Slide-tags data—a technology combining Takara Bio Trekker spatial barcoding, 10x Genomics single-nucleus capture, and Illumina sequencing—to achieve ~3.5 µm spatial resolution. The computational workflow transforms raw FASTQ files into spatially-resolved gene expression maps using a four-stage R/Python/Shell pipeline involving DBSCAN clustering and UMI-weighted centroids.
Complementing the research tools is a "Job Intelligence System," a 10-step automated pipeline that scrapes job postings from Roche and Novartis. It utilizes Google’s Gemini API to perform dual-stage scoring: evaluating domain relevance to spatial genomics and matching positions to a specific candidate profile. The repository creates a strategic feedback loop where technical research outputs (e.g., Voronoi visualizations, Scanpy/Squidpy objects) serve as portfolio evidence, while AI-identified skill gaps in high-relevance job postings guide subsequent research and technical development.
Strategic Overview: Integrated Spatial Genomics and Career Intelligence
Slide-tags Technology Foundation: The system integrates three distinct technologies (Takara Bio, 10x Genomics, and Illumina) to produce spatially-resolved single-nucleus gene expression maps at 3.5 µm resolution.
Takeaway: This high-resolution mapping maintains single-cell transcriptomic quality by utilizing UV-mediated release of DNA barcodes into tissue sections.
Data Processing Pipeline Architecture: A four-stage transformation sequence:
sb_processing.sh: Filters and downsamples raw FASTQ files using anchor sequences.
cell_barcode_matcher.R: Extracts spatial barcode (SB) and cell barcode (CB) pairings.
bead_matching.py: Performs fuzzy barcode matching to associate SBs with specific (x,y) coordinates.
spatial_positioning.R: Uses DBSCAN clustering to identify primary nucleus locations and calculates UMI-weighted centroids.
Job Intelligence System (Stages 1–10): A comprehensive pipeline that transforms unstructured web data from Roche/Novartis career sites into structured, AI-scored reports.
Takeaway: The system uses Selenium and Playwright for acquisition, SQLite for persistence, and Gemini-3-Flash for intelligent evaluation.
Dual-Stage AI Scoring: Jobs are evaluated on two 1-5 scales:
Slide-tag Relevance: Measures the job's alignment with spatial transcriptomics technology.
Candidate Match: Compares the job summary against a candidate_profile.md to determine fit based on skills and experience.
Roche Corporate Intelligence: The repository tracks Roche’s business units, specifically the Pathology Lab (Diagnostics) and Molecular Lab, identifying them as the primary commercial homes for spatial technologies.
Important Detail: Roche's "Sequencing by Expansion" (SBX) platform, launching in 2026, is identified as a critical technology window for spatial genomics integration due to its high throughput (5B reads/hour).
Brain Analysis Visualization System: A Python-based visualization suite that processes .h5ad objects to generate Voronoi tessellations and UMAP plots.
Takeaway: These artifacts serve as "Portfolio Evidence," demonstrating proficiency in Scanpy, Squidpy, and computational geometry to potential employers.
API Efficiency and Chunking: To manage LLM token limits, the system implements a 5,100-word limit per API request and a 3-attempt retry mechanism for resilience.
Takeaway: This architecture allows for efficient batch processing of 10–20 jobs per request while ensuring data persistence via CSV checkpointing.
Strategic Feedback Loop: The system identifies "Skill Gaps" by comparing high-scoring job requirements against the candidate profile.
Takeaway: If a high-match job requires a specific skill (e.g., Nextflow), the workflow dictates developing a research project to add that skill to the portfolio, thereby increasing future match scores.
Multi-Format Reporting: Scored results are exported into Markdown, LaTeX, and Typst formats to facilitate both quick review and professional PDF generation for formal applications.