Repository navigation
Commit 0de3096
Align the training-cell loadings with the rows they are regressed against
Fixes #546.
`_bicv_trial` built the held-out gene regression's design matrix by
concatenating per-condition blocks in condition order:
Z = np.concatenate(
[(P_train[i] @ B) * A[i] for i in range(n_cond) if np.any(cond_train == i)],
axis=0,
)
but regressed it against `X_train_test_genes`, which stays in the training
cells' natural row order. `P_train[i]` lists condition `i`'s cells in their
within-condition order, so concatenating produces `Z` sorted by condition. The
two orderings coincide only when conditions happen to be stored as contiguous
blocks. On pooled data -- a Perturb-seq screen, or anything not physically
sorted by sample before sequencing -- conditions are interleaved, and row `k` of
`Z` then describes a different cell from row `k` of the expression it is paired
with. Every held-out R2X the function reports is computed from that
least-squares fit.
`_train_cell_loadings` scatters each condition's block back to the positions its
cells actually occupy. Same asymptotic cost, and identical output in the
contiguous case.
Three of the four new tests fail against the previous implementation and pass
against this one; the fourth asserts the no-op on contiguous input, and passes
either way by construction. That split is deliberate -- without the
"differ when interleaved" test, an implementation that quietly reverted to
concatenation would still satisfy the rest.
The missing-condition case is also covered: when a fold leaves some condition
with no training cells, concatenation produced the right number of rows in the
wrong order, which is the same defect in a form that is even harder to notice.
Why every existing test missed this: `_make_test_data` builds conditions as
contiguous blocks, which is precisely the case where the bug is invisible. The
end-to-end test added here stores the same cells interleaved.
Reported by @aarmey, with the diagnosis and the suggested fix.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>1 parent 0d344ac commit 0de3096
2 files changed
Lines changed: 145 additions & 5 deletions
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
76 | 76 | | |
77 | 77 | | |
78 | 78 | | |
| 79 | + | |
| 80 | + | |
| 81 | + | |
| 82 | + | |
| 83 | + | |
| 84 | + | |
| 85 | + | |
| 86 | + | |
| 87 | + | |
| 88 | + | |
| 89 | + | |
| 90 | + | |
| 91 | + | |
| 92 | + | |
| 93 | + | |
| 94 | + | |
| 95 | + | |
| 96 | + | |
| 97 | + | |
| 98 | + | |
| 99 | + | |
| 100 | + | |
| 101 | + | |
| 102 | + | |
| 103 | + | |
79 | 104 | | |
80 | 105 | | |
81 | 106 | | |
| |||
127 | 152 | | |
128 | 153 | | |
129 | 154 | | |
130 | | - | |
131 | | - | |
132 | | - | |
133 | | - | |
| 155 | + | |
134 | 156 | | |
135 | 157 | | |
136 | 158 | | |
| |||
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
7 | 7 | | |
8 | 8 | | |
9 | 9 | | |
10 | | - | |
| 10 | + | |
11 | 11 | | |
12 | 12 | | |
13 | 13 | | |
| |||
150 | 150 | | |
151 | 151 | | |
152 | 152 | | |
| 153 | + | |
| 154 | + | |
| 155 | + | |
| 156 | + | |
| 157 | + | |
| 158 | + | |
| 159 | + | |
| 160 | + | |
| 161 | + | |
| 162 | + | |
| 163 | + | |
| 164 | + | |
| 165 | + | |
| 166 | + | |
| 167 | + | |
| 168 | + | |
| 169 | + | |
| 170 | + | |
| 171 | + | |
| 172 | + | |
| 173 | + | |
| 174 | + | |
| 175 | + | |
| 176 | + | |
| 177 | + | |
| 178 | + | |
| 179 | + | |
| 180 | + | |
| 181 | + | |
| 182 | + | |
| 183 | + | |
| 184 | + | |
| 185 | + | |
| 186 | + | |
| 187 | + | |
| 188 | + | |
| 189 | + | |
| 190 | + | |
| 191 | + | |
| 192 | + | |
| 193 | + | |
| 194 | + | |
| 195 | + | |
| 196 | + | |
| 197 | + | |
| 198 | + | |
| 199 | + | |
| 200 | + | |
| 201 | + | |
| 202 | + | |
| 203 | + | |
| 204 | + | |
| 205 | + | |
| 206 | + | |
| 207 | + | |
| 208 | + | |
| 209 | + | |
| 210 | + | |
| 211 | + | |
| 212 | + | |
| 213 | + | |
| 214 | + | |
| 215 | + | |
| 216 | + | |
| 217 | + | |
| 218 | + | |
| 219 | + | |
| 220 | + | |
| 221 | + | |
| 222 | + | |
| 223 | + | |
| 224 | + | |
| 225 | + | |
| 226 | + | |
| 227 | + | |
| 228 | + | |
| 229 | + | |
| 230 | + | |
| 231 | + | |
| 232 | + | |
| 233 | + | |
| 234 | + | |
| 235 | + | |
| 236 | + | |
| 237 | + | |
| 238 | + | |
| 239 | + | |
| 240 | + | |
| 241 | + | |
| 242 | + | |
| 243 | + | |
| 244 | + | |
| 245 | + | |
| 246 | + | |
| 247 | + | |
| 248 | + | |
| 249 | + | |
| 250 | + | |
| 251 | + | |
| 252 | + | |
| 253 | + | |
| 254 | + | |
| 255 | + | |
| 256 | + | |
| 257 | + | |
| 258 | + | |
| 259 | + | |
| 260 | + | |
| 261 | + | |
| 262 | + | |
| 263 | + | |
| 264 | + | |
| 265 | + | |
| 266 | + | |
| 267 | + | |
| 268 | + | |
| 269 | + | |
| 270 | + | |
0 commit comments