Pairwise differences
Positive logFC means higher in the comparison condition than the reference. An overall condition test asks whether any condition differs; its result has no single direction.
Every test keeps its full table. A result with no significant features is complete. Tests without enough evidence are labeled separately. FDR is calculated within each test; counts of discoveries across methods or comparisons must not be added together.
Coverage in each search
Input features and features retained for primary protein and supporting peptide testing. Gene comparisons are in Optional analyses. DDA and DDA+ are separate searches of the same samples.
| Cohort | Search | Evidence | Input features | Retained for testing | Status |
|---|
No main statistical coverage summary was generated. Count matrices and optional results are available in their sections.
| Question / comparison | Evidence | Cohort / search | Method | Tested / reported | Significant | Status | Result tables |
|---|
No tests match these filters.
Peptide and isoform evidence
Sequence mapping and detection summaries are descriptive evidence. Their statistical comparisons are listed above when selected.
Counts, matrices and study quality checks
Files are grouped by cohort and search mode. Raw counts preserve observed evidence; normalized values support comparisons, and z-scores help display patterns.
Modification-by-gene/protein counts measure distinct modified peptide sequences per modification type. Gene counts come from proteomics. Z-scores are visualization values and have no statistical significance threshold.
Reading your results
Drug interaction
interaction_log2FC = combination − A − B + control. Positive values are above the multiplicative expectation; negative values are below it. This measures proteomic nonadditivity. Cell-viability synergy needs functional evidence.
Adjusted p-values
FDR is the Benjamini–Hochberg adjusted p-value within each method, cohort, search mode, feature type and comparison. Each table uses its configured cutoff. Choosing the smallest p-value across methods is not a combined test.
Non-significant and insufficient-evidence results
“No significant results” means tests were computed but none passed the FDR cutoff. “Insufficient evidence” means no estimable test was available. “No eligible evidence” means there were no eligible feature rows. Untestable rows have blank p-values/FDR and an explicit status; they do not establish absence of an effect. Header-only significant tables are valid.
Counts, condition means and detection
Protein and peptide matrices contain spectral counts; optional gene matrices use unambiguous peptide-to-gene assignments. Exclusion of shared peptides applies only to gene counts; protein counts and their statistical power are unchanged. Gene tests use the DEPs settings with separate models and FDR. Modification matrices count distinct modified peptide sequences per entity and modification type. These measure detection evidence, not modification occupancy.
Condition means include zero-count samples. Detection columns show samples with observed counts out of the total samples in each condition. Detection alone does not establish a significant difference or biological absence.
Main standard counts use TMM normalization. Sparse modification and optional peptide/isoform comparisons use TMMwsp. Normalization is separate for each cohort, search mode and evidence type. Z-scores standardize log2 CPM within each feature; they are for visualization.
Methods and Monte Carlo precision
edgeR uses count-based quasi-likelihood models. limma-voom uses precision-weighted log2 counts and empirical Bayes moderation. Optional peptide/isoform and modification comparisons use the edgeR small-count model. edgeRun adds pairwise exact tests where selected.
Monte Carlo drug interaction simulates negative-binomial counts from a fitted model with both main drug effects and zero interaction. Fitted dispersion and normalization stay fixed, making this an approximate parametric bootstrap. Both coefficient models are refitted in each simulation. The empirical p-value is (exceedances + 1) / (simulations + 1).
The smallest possible Monte Carlo p-value is 1 / (simulations + 1), which can limit FDR discoveries. Its uncertainty intervals describe simulation precision, not biological effect uncertainty. Limma interaction tables include effect confidence intervals.
Isoforms, HLA peptides and the study design
Isoform evidence requires peptides that distinguish an accession from alternatives in the searched FASTA, allowing I/L equivalence. Counts without this support do not receive inferred isoform-specific tests. Isoform-associated evidence does not establish isoform switching. HLA peptide evidence does not measure source-protein expression.
Biological replicates are assumed independent; batch and paired-subject effects are not inferred. DDA and DDA+ remain separate searches. Unconfirmed matching of drug doses and treatment times is labeled an exploratory combination contrast in the tables.
Methods and supporting files
Result folder guide · Complete result index · Model records and file checksums
Model inputs, filtering masks, fitted designs, normalization factors, settings and software versions are kept under analysis_details. The final result layout preserves the statistical tables without refitting.