Table 1.
Significant and unique HOSVD combinations, that is, subtensors
| Subtensor | Overexpression | Underexpression | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Fraction (%) | Annotation | J | j | P-value | Annotation | J | j | P-value | ||
| 1 | 1, 1, 1 | 72.3 | — | — | — | — | — | — | — | — |
| 2 | 2, 2, 1 | 8.7 | M/G1 | 84 | 41 | 1.5 × 10−33 | S/G2 | 88 | 27 | 6.3 × 10−16 |
| 3 | 3, 3, 1 | 6.7 | G1 and S | 263 | 103 | 1.1 × 10−77 | G2/M | 141 | 53 | 2.6 × 10−36 |
| 4 | 4, 1, 2 | 2.7 | 3′-end ARS | 153 | 14 | 1.1 × 10−2 | Histone genes | 9 | 9 | 9.1 × 10−13 |
| 5 | 5+6, 1, 3 | 1.9 | Histone genes | 9 | 7 | 1.5 × 10−8 | 3′-end ARS | 153 | 16 | 1.9 × 10−3 |
| 6 | 7, 3, 2 | 0.8 | Histone genes | 9 | 4 | 4.9 × 10−4 | — | — | — | — |
| 7 | 8, 3, 3 | 0.7 | — | — | — | — | 3′-end ARS | 153 | 17 | 6.9 × 10−4 |
The fractions of mRNA expression that the seven most significant combinations capture in the data cuboid (Figure 1), and the probabilistic significance of the enrichment of the corresponding patterns of expression variation across the genes in overexpressed or underexpressed cell cycle-regulated genes (Spellman et al, 1998), histone genes and genes with ARSs near their 3′ ends (Cherry et al, 1997; Nieduszynski et al, 2007). The P-value of each enrichment is calculated (Supplementary information Section 2.5) according to the annotations of the genes (Supplementary information Datasets 4 and 5), assuming hypergeometric distribution of the J annotations among the K=4270 genes, and of the subset of j⊆J annotations among the subset of k=200 genes with largest or smallest levels of expression in the corresponding pattern (Supplementary information Dataset 6), as described (Tavazoie et al, 1999).