USP1 does not behave like part of the Fanconi module, and does not track ID1 either
also called USP1 module membership · USP1 co-dependency · USP1 neither home
We thought USP1 might belong to the repair module our scan implicated, or else to a rival pathway. Across 1,208 cell lines it behaves like neither, which means we should stop counting it as evidence for the first one.
The same thing, for a scientist
Co-dependency across **1,208 DepMap lines**. If USP1 acted through the FA monoubiquitination module its profile should track FANCL, UBE2T, FANCD2, FANCI, FANCA and FANCF; if through ID1, it should track ID1. **It tracks neither.** USP1 against the module runs **r = +0.098 to +0.207**, against a **module internal median of r = 0.576** over 15 pairs. USP1~ID1 is **−0.022**. The negative control USP1~ACTB is **−0.127**, larger in magnitude than several of the module correlations.
The numbers, and the yardstick that makes them readable
| pair | Pearson r | Spearman rho |
|---|---|---|
| USP1 ~ FANCL | +0.207 | +0.193 |
| USP1 ~ UBE2T | +0.116 | +0.127 |
| USP1 ~ FANCD2 | +0.098 | +0.099 |
| USP1 ~ FANCI | +0.147 | +0.123 |
| USP1 ~ FANCA | +0.162 | +0.128 |
| USP1 ~ FANCF | +0.161 | +0.140 |
| USP1 ~ ID1 | -0.022 | -0.022 |
| USP1 ~ ATR | +0.067 | +0.109 |
| USP1 ~ RPA1 | -0.012 | +0.005 |
| USP1 ~ ACTB (control) | -0.127 | -0.120 |
The yardstick is the module's coherence with itself: median Pearson r = 0.576 across its 15 internal pairs. FANCL, UBE2T, FANCD2, FANCI, FANCA and FANCF behave like one unit. USP1 does not join it — its strongest link, to FANCL at +0.207, is a third of the module's internal median.
And it does not join the alternative. USP1~ID1 is −0.022, which is nothing.
The control is the part that sets the floor. USP1ACTB comes out at −0.127 — bigger in absolute size than USP1FANCD2 (+0.098) or USP1~UBE2T (+0.116). Correlations of that magnitude in this matrix are not evidence of anything; they are what two unrelated genes look like.
What this costs the FA reading, and what it does not
It costs USP1. ana-fa-pathway-enrichment-test listed USP1 at rank 129 among the module members carrying the enrichment, and that was the most rhetorically useful member — the E2, the E3, both substrates and the deubiquitinase that reverses them, all near the top. The deubiquitinase should come off that list. Its selectivity in KMT2A-rearranged lines is real and is not explained by FA-module membership.
It does not cost the module. FANCL, UBE2T, FANCD2, FANCI, FANCA and FANCF cohere at r = 0.576 and all rank in the top 500 of 17,728. Removing USP1 removes one member from a list of eight, and the enrichment statistic was computed over the whole Reactome pathway rather than over that hand-picked subset, so p = 5.6 × 10⁻⁸ does not move.
It also does not rescue the ID1 story. The published leukaemia mechanism for USP1 is ID1/AKT, and the dependency data give USP1~ID1 = −0.022. Whatever makes USP1 selectively required in these lines, the co-dependency structure does not point at either of its two known substrate axes.
Lineage-split, because this project's own rule says it must be
Standing rule 7: a DepMap dependency claim must be LINEAGE-SPLIT before it is stated. The first version of this analysis ran over all 1,208 lines and stated a conclusion, which is exactly what the rule forbids. Re-run in three strata:
| stratum | n lines | module internal median | USP1 best module r | USP1~ID1 | USP1~ACTB |
|---|---|---|---|---|---|
| all lines | 1208 | +0.576 | +0.207 | -0.022 | -0.127 |
| haematological | 63 | +0.649 | +0.343 | +0.013 | -0.059 |
| other lineages | 1145 | +0.546 | +0.217 | -0.024 | -0.131 |
The conclusion survives. In haematological lines USP1's best correlation with the module rises to +0.343 — but the module's coherence with itself rises too, to +0.649. The ratio barely moves: USP1 sits at roughly half the module's internal coherence in every stratum (0.36, 0.53, 0.40). And USP1~ID1 stays at nothing everywhere (+0.013 in haematological).
A conclusion that flips between strata is a different finding from one that holds. This one holds — which is the outcome the rule exists to distinguish, and it was worth one API call to know rather than assume.
- Lineage split done and the conclusion holds, but the haematological stratum is 63 lines and USP1's module correlation rose from +0.207 to +0.343 in it. A larger haematological panel is the test that could still overturn this.
- Test USP1 against PCNA and the translesion-synthesis axis - its third substrate - since 'neither home' may mean the right home was not tested.
- Does USP1's KMT2A-rearranged selectivity survive at all once lineage is conditioned on? Its rank of 129 came from the eleven-line panel, seven of whose nine patients are myeloid.
- Re-examine the other module members the same way. USP1 was tested because it was the druggable one, not because it was the doubtful one, and FAAP100 or SLX4 might fail the same test.
- E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.analysis
Across 1,208 DepMap cell lines, USP1's CRISPR gene-effect profile correlates with the FA monoubiquitination module at Pearson r = +0.098 to +0.207 (FANCL +0.207, FANCA +0.162, FANCF +0.161, FANCI +0.147, UBE2T +0.116, FANCD2 +0.098), against a module internal median of r = 0.576 over 15 pairs. USP1 against ID1 is -0.022, against ATR +0.067, against RPA1 -0.012, and against the negative control ACTB -0.127.
USP1 module membership: does its cross-line dependency profile track the Fanconi monoubiquitination module or the ID1 axis? · Oncofusion knowledge graph (internal analysis) 2026What this evidence cannot show
Fetched from the Open Targets GraphQL API on 2026-09-20; 1,208 lines shared across all pairs, so no missing-data asymmetry. Co-dependency is correlation across cell lines and is computed over ALL lineages here, not restricted to leukaemia — a restricted version could differ and was not run. The DepMap release version is not stated by the API.
Evidence that argues against this (2)
- Weak co-dependency does not exclude pathway membership. A deubiquitinase acts opposite to the ligase it counters, so USP1 and FANCL could be in one pathway and correlate weakly or negatively by construction - losing the eraser is not equivalent to losing the writer. The module's internal coherence is between components acting in the same direction, which is an unfair yardstick for the one component that acts against them. Oncofusion knowledge graph (internal analysis) 2026
- USP1 is a promiscuous deubiquitinase with substrates beyond FANCD2 and ID1, including PCNA-Ub in translesion synthesis. 'Neither home' may simply mean the relevant home was not among the two tested, and this analysis cannot distinguish that from USP1's selectivity being an artefact. Open Targets 2026
- E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.analysis
The FA monoubiquitination module coheres as a unit in the same data: FANCL, UBE2T, FANCD2, FANCI, FANCA and FANCF correlate with each other at a median Pearson r of 0.576 across 15 pairs, which is the yardstick against which USP1's 0.098-0.207 is read.
USP1 module membership: does its cross-line dependency profile track the Fanconi monoubiquitination module or the ID1 axis? · Oncofusion knowledge graph (internal analysis) 2026What this evidence cannot show
Same fetch and same 1,208 lines. A high internal correlation among FA genes is expected and is partly a statement that the annotation is right rather than a discovery; it is used here as a scale, not as a finding.
Evidence that argues against this (1)
- Strong internal co-dependency within a DNA-repair module is also what a shared response to screen-level stress would look like - all six genes buffering the same insult in the same lines. The coherence validates the yardstick without establishing that the module's selectivity in KMT2A-rearranged lines is fusion-driven. Oncofusion knowledge graph (internal analysis) 2026
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.analysis
The 'neither home' conclusion survives a lineage split. Across 63 haematological lines USP1's best correlation with the FA module is +0.343 against a module internal median of +0.649; across 1,145 other lineages it is +0.217 against +0.546; across all 1,208 it is +0.207 against +0.576. USP1~ID1 is +0.013, -0.024 and -0.022 respectively. In every stratum USP1 sits at roughly half the module's internal coherence and shows no relationship with ID1.
USP1 module membership, lineage-split: haematological against other lineages · Oncofusion knowledge graph (internal analysis) 2026What this evidence cannot show
Run 2026-09-20 on the same Open Targets fetch. **The haematological stratum is only 63 lines**, so +0.343 and +0.649 both carry wide uncertainty and no confidence interval was computed. Lineage was assigned by string-matching the screen's diseaseFromSource field for leukaemia, lymphoma, myeloma and myelodysplasia, which is coarser than Cellosaurus annotation and will misfile some lines.
Evidence that argues against this (2)
- The ratio being stable across strata is reassuring about the conclusion and says nothing about the mechanism. USP1 could correlate at half the module's coherence in every lineage precisely because it is a downstream eraser acting against them, which would make the stability an argument FOR membership rather than against it. The split tests robustness, not interpretation. Oncofusion knowledge graph (internal analysis) 2026
- Sixty-three haematological lines is a thin stratum for a correlation, and the one number that moved most - USP1's best module correlation, +0.207 to +0.343 - moved in the direction that would weaken this entry's conclusion. With a larger haematological panel that gap could close. Oncofusion knowledge graph (internal analysis) 2026