Avasthā-Reader: reference distribution and code for amino-acid-derived Tridoṣa reading of bulk transcriptomes

# Avasthā-Reader: reproducibility deposit This archive contains the reference data and the core computational codeunderlying: > Pande, A. *Avasthā-Reader: an amino-acid-derived Tridoṣa reading> distinguishes cancer stages across six epithelial transcriptomes.*> Manuscript under review. It is the reproducibility record for the per-sample Tridoṣa readings andclass-similarity scores reported in that work. ## What Avasthā-Reader is Avasthā-Reader is a **training-free, similarity-based reader** of intrinsiccompositional structure in bulk transcriptomes. For a bulk gene-expressionvector it computes three compositional fractions of the expressed proteome —**mobility (Vāta)**, **transformation (Pitta)**, **stability (Kapha)** — byexpression-weighted aggregation over per-gene Tridoṣa values, and reportsthe sample's position relative to a reference distribution of **7,611 bulkhuman transcriptomes** spanning six epithelial cancers (colorectal, breast,liver, lung, thyroid, kidney) from TCGA pan-cancer and ARCHS4 v2.5. It is a research instrument for hypothesis generation. It returns continuouscompositional values and *relative* class-similarity scores; it does **not**return a categorical diagnostic verdict, and is **not** a clinical tool. An interactive deployment is hosted at:https://huggingface.co/spaces/amitpande74/avastha-reader ## Contents ```avastha_reader_deposit/├── README.md this file├── code/│ ├── tridosha_compute.py per-sample Tridoṣa computation│ ├── similarity_scorer.py Mahalanobis class-similarity scoring│ └── run_avastha_reader.py end-to-end command-line interface├── data/│ ├── gene_prakriti.tsv per-gene Tridoṣa table (20,210 genes, 1.3 MB)│ ├── reference_distribution.tsv reference distribution (7,611 samples, 844 KB)│ └── class_centroids.tsv precomputed centroids (21 rows, 4 KB)└── examples/ └── example_input.tsv format template (2,400 genes)