← Insights From BioData
Computed workflow · Human-GEM 1.12.0

From synthetic expression to reaction activity

COBAMP GPR scoring and TROPPO iMAT with Gurobi across 200 artificial samples. Three staged inputs, one verified model, complete reaction matrices and auditable output tables.

Synthetic data only · no biological interpretation
Explore the figuresDownload results

Three layers, one cohort

The same 100 AD-labelled and 100 control-labelled synthetic samples pass through each layer. Only AD-labelled expression is modified. Figures are calculated from the supplied inputs, committed reaction matrices, and downstream statistics.

Layer 01

Expression only

6,848 median active reactions / AD-labelled sample

5290 Fisher raw p < .0517 reporter corrected p < .05
Sample diagnostics ↗

Layer 02

Splicing informed

6,832 median active reactions / AD-labelled sample

5322 Fisher raw p < .0517 reporter corrected p < .05
Sample diagnostics ↗

Layer 03

Splicing + pathogenicity

6,794 median active reactions / AD-labelled sample

5356 Fisher raw p < .0518 reporter corrected p < .05
Sample diagnostics ↗

Four views of the computed run

Each plot is generated by 05_plot_results.py from the solved matrices, the supplied target list, and calculated statistics. Dots in the reaction-count figures represent individual synthetic samples.

Active reaction count distributions for controls and three AD-labelled model stages
Controls are shown once because their input and reaction vector are unchanged. Each group or stage contains 100 samples.
Within-sample changes in active reaction counts after splicing and pathogenicity adjustments
Paired AD models. Baseline minus splicing median: 0 reactions; IQR -22–47; 0 Tukey outliers. Positive means fewer active reactions after adjustment.
Expression categories of provided splicing targets before zeroing
800 supplied gene–sample targets across 100 AD-labelled samples; 774 changed from a nonzero value. Categories use the documented low and high cutoffs at gene level. A complete splice-event detection table was not supplied.
Comparison of corrected reporter-like pathway statistics from baseline and splicing-adjusted models
Only pathways passing the archived reporter-like corrected p < 0.05 in either contrast appear. Correction is the archived permutation normalization (1,000 permutations, fixed seed), not Benjamini–Hochberg.

Pairing and overlap checks

All 100 AD-labelled samples were matched to their own adjusted model. Median active-reaction Jaccard: 0.923; median Overlap index: 0.963. Significant reporter-like pathway-set overlap: 0.700 Jaccard / 0.824 Overlap. The pathway comparison is undefined when either set is empty.

Per-sample similarity and model-size data · Target expression audit

Audit the statistics

Reaction-level Fisher p-values and reporter-like pathway p-values answer different questions. An associated subsystem below is not necessarily a statistically significant pathway.

Lowest Fisher p-values

LayerReactionRaw pBH p
Expression onlyMAR041222.2e-597.7e-58
Expression onlyMAR044972.2e-597.7e-58
Expression onlyMAR043512.2e-597.7e-58
Splicing informedMAR041222.2e-599.5e-58
Splicing informedMAR044972.2e-599.5e-58
Splicing informedMAR043512.2e-599.5e-58
Splicing + pathogenicityMAR041222.2e-591.3e-57
Splicing + pathogenicityMAR044972.2e-591.3e-57
Splicing + pathogenicityMAR043512.2e-591.3e-57

Ranked reporter-like rows

LayerSubsystemCorrected p
Expression onlyTransport reactions5e-13
Expression onlyFatty acid biosynthesis (odd-chain)3.2e-07
Expression onlyHeparan sulfate degradation3.6e-06
Splicing informedTransport reactions7.1e-16
Splicing informedFatty acid biosynthesis (odd-chain)4e-07
Splicing informedHeparan sulfate degradation7.9e-06
Splicing + pathogenicityTransport reactions1.8e-18
Splicing + pathogenicityFatty acid biosynthesis (odd-chain)1.8e-07
Splicing + pathogenicityHeparan sulfate degradation2.2e-06

Subsystems associated with raw Fisher p < .05 reactions

LayerAssociated subsystemReaction countReporter significant?
Expression onlyTransport reactions2051Yes
Expression onlyExchange/demand reactions546No
Expression onlyDrug metabolism286Yes
Splicing informedTransport reactions2047Yes
Splicing informedExchange/demand reactions551No
Splicing informedDrug metabolism287Yes
Splicing + pathogenicityTransport reactions2046Yes
Splicing + pathogenicityExchange/demand reactions552No
Splicing + pathogenicityDrug metabolism288Yes

Download the complete association table

Download and reproduce

The complete matrices are gzip-compressed TSV files. Decompress them with gzip -dc file.tsv.gz. The accompanying diagnostics and statistics are ordinary TSV files.

Expression only

Reaction matrix · 13,070 × 200Sample diagnosticsAll Fisher testsReporter-like results

Splicing informed

Reaction matrix · 13,070 × 200Sample diagnosticsAll Fisher testsReporter-like results

Splicing + pathogenicity

Reaction matrix · 13,070 × 200Sample diagnosticsAll Fisher testsReporter-like results

Analysis summary TSV · SHA-256 manifest · Run metadata · Run instructions

Interpretation boundary. These are artificial inputs and synthetic scenario labels. No plotted difference is evidence about Alzheimer disease, diagnosis, or treatment. Mixed-integer models can have alternative optimal flux solutions; the committed Gurobi matrices are the reference for reproducing these figures.