Supplementary material for

  NR5A1 consecutive-exon skipping in human adrenal gland, ovary, pituitary and
  testis: SRSF6 and CLK4 track it in all but testis
  Kenji Ohe

Every script here queries only public endpoints (Snaptron at
snaptron.cs.jhu.edu, the GTEx annotation files at storage.googleapis.com) or
reads a table shipped alongside it. None requires credentials. They were run
on Python 3.13 with numpy, pandas, scipy, statsmodels and matplotlib.

Each script takes an optional directory argument, which defaults to the
current directory, and both reads its inputs from and writes its outputs to
that directory:

  python3 SupplementaryFileS3_normalisation.py .


ANALYSIS SCRIPTS
----------------

File S1  SupplementaryFileS1_splicing_factors.py
    Retrieves NR5A1 junction counts and the mRNA abundance of sixteen splicing
    regulators across the four tissues, removes duplicated GTEx sample entries,
    and runs the within-tissue rank partial correlations and the signed
    Stouffer combination reported in Table 2.

File S2  SupplementaryFileS2_global_and_steroid.py
    Builds the global skipping index from the annotated junctions of 48 broadly
    expressed genes, and runs the steroidogenic dosage-bound tests.
    Writes Table S1.

File S3  SupplementaryFileS3_normalisation.py
    Repeats the twelve factor-by-tissue tests under five normalisers and three
    models, and tests each normaliser against the skip index itself.
    Input: sample_quality_input.csv. Writes Table S2 and, as
    normaliser_diagnostics.csv, the per-normaliser diagnostics plotted in
    Figure 2A.

File S4  SupplementaryFileS4_ovary_and_zonation.py
    Ovary: donor age, menopausal band, and follicular activity as recorded in
    the GTEx pathology notes, with the agreement between the pathology
    classification and the follicle marker genes. Adrenal gland: the cortical
    zonation scores under three nested covariate sets, and the test of whether
    NR4A2 tracks the Hardy death classification rather than zone size.
    Input: sample_quality_input.csv. Writes ovary_age_pathology.csv and
    adrenal_zonation.csv.

File S5  SupplementaryFileS5_sra_discovery.py
    Searches the Snaptron srav3h compilation for public runs carrying the NR5A1
    junctions, classifies them from their metadata, and performs the
    depth-matched comparisons. Writes Table S3.

File S6  SupplementaryFileS6_figures_1_3_4.py
    Draws Figures 1, 3 and 4 from Table S1. Writes Tables S5 and S6.
    Requires no network access.

File S7  SupplementaryFileS7_figure_2.py
    Draws Figure 2 from Table S2. Writes Table S4. Requires no network access.


DATA TABLES
-----------

Table S1  SupplementaryTableS1_per_sample.csv            1,054 rows
    One row per GTEx sample, after duplicate removal: rail identifier, external
    identifier, tissue, exon 6->7 denominator reads, skipped reads, skip index,
    the three sample-quality variables, the housekeeping total, and the
    abundance of NR5A1, CLK2, CLK4, SRSF6 and ten steroidogenic genes relative
    to that total, with the global skipping index.
    The column gsi_topread is the global skipping index used throughout. The
    column gsi is a variant taking the shortest junction of each group as the
    reference; it gives medians of 37-56%, indicating that the shortest
    junction is often not the constitutive one, and it is NOT used anywhere in
    the manuscript.

Table S2  SupplementaryTableS2_normalisation.csv           225 rows
    The 180 factor-by-tissue-by-scheme tests (three factors x four tissues x
    five normalisers x three models) with r, P and n, plus the 45 signed
    Stouffer combinations. Rows are typed by the `analysis` column.

Table S3  SupplementaryTableS3_sra_runs.csv              1,436 rows
    Public SRA runs carrying at least 20 reads on the exon 6->7 junction, with
    their junction counts, skip index, study accession and the metadata fields
    used to classify them.

Table S4  SupplementaryTableS4_figure2_values.csv           12 rows
    For each factor and tissue: the primary-scheme correlation, the median,
    the minimum and maximum across the fifteen schemes, and the number of
    schemes reaching P < 0.05. These are the values drawn in Figure 2B.

Table S5  SupplementaryTableS5_figure3_values.csv           12 rows
    For each factor and tissue: the correlation with the global skipping index,
    with the NR5A1 skip index, and with the NR5A1 index given the global index,
    each with its P value and 95% confidence interval. Figure 3.

Table S6  SupplementaryTableS6_figure4_values.csv           80 rows
    Percentage change in each steroidogenic transcript per percentage point of
    skip index, by tissue, unadjusted and adjusted for CLK2, CLK4 and SRSF6,
    with 95% confidence intervals and P values. Figure 4.


INPUT
-----

sample_quality_input.csv
    The per-sample skip index and sample-quality variables for the 1,054
    samples, used as the input to Files S3 and S4. It is the same sample set as
    Table S1, carried separately so that those two scripts can be run on their
    own.


FIGURES
-------

figures/Fig1_specificity.pdf        Figure 1, from File S6
figures/Fig2_normalisation.pdf      Figure 2, from File S7
figures/Fig3_genomewide.pdf         Figure 3, from File S6
figures/Fig4_dosage_bound.pdf       Figure 4, from File S6

Tissue colours are the Okabe-Ito set: adrenal gland #0072B2, ovary #009E73,
pituitary #E69F00, testis #CC79A7. Every adjacent pair clears a
colour-vision-deficiency separation check, and tissue is named on every axis,
so no panel depends on colour alone.
