AI-Augmented Independent Research as a New Scientific Methodology: Four Computational Studies in Systemic Sclerosis

Abstract

This dissertation presents four computational studies in systemic sclerosis (SSc), a rare autoimmune disease characterized by progressive skin fibrosis, vascular damage, and immune dysregulation. The studies were conducted by a single independent researcher operating outside any institutional affiliation, using publicly available genomic datasets and an AI-augmented research methodology developed between July 2025 and February 2026. Paper 1: Single-cell drug repurposing for scleroderma skin using scVI and LINCS reversal. Identifies MEK inhibitor PD-0325901 and SRC/ABL inhibitor dasatinib as candidate therapeutics for reversing fibroblast activation. Under review at Scientific Reports. Paper 2: Single-cell signatures of EBV-associated immune scarring in SSc B cells. Demonstrates that EBV immune-scarring transcriptional programs characterize the majority of circulating SSc B cells. BTK inhibitors identified as top repurposing candidates. Paper 3: T cell functional states and regulatory dynamics in SSc peripheral blood. Recovers coherent Th1, Th2, Th17, Tfh, Treg, and cytotoxic T cell populations using program-level scoring. Under review at Frontiers in Immunology. Paper 4: B cell activation associates with reduced pro-fibrotic T cell programs despite amplified inflammatory ligand expression. Proposes Galectin-9/TIM-3 immunoregulatory feedback as a mechanism explaining inconsistent rituximab results in SSc clinical trials. The dissertation also contributes a methodology chapter documenting the AI-Augmented Independent Research (AAIR) framework: a transparent, replicable approach to conducting computational biology research without institutional infrastructure, using large language models as primary research and coding partners. AI co-authors (ChatGPT-4, Claude, Gemini Pro) are named in recognition of their substantive contributions to code generation, manuscript drafting, and hypothesis development. This work is submitted as a proof of concept arguing that research institutions should develop frameworks for evaluating and credentialing AI-augmented independent scholarship. All code is archived at github.com/glenritchel/scleroderma-scvi and doi.org/10.5281/zenodo.17281987. Data: GSE138669, GSE195452, GSE210395 (NCBI GEO).

Ähnliche Arbeiten