Table 4.

Chi-square analysis for association of gene expression with subtypes

	Basal-like	HER2+/ER-	Luminal A	Luminal B	Normal-like	p-value
# tumors	53	35	123	55	29
Cluster 1a	68%	37%	12%	56%	14%	<0.0001
Cluster 2a	89%	49%	5%	49%	7%	<0.0001
Cluster 3a	77%	51%	11%	47%	0%	<0.0001
EGFRa	68%	20%	27%	18%	41%	<0.0001
HER2a	15%	100%	28%	26%	24%	<0.0001
HER4*	9%	3%	50%	38%	31%	<0.0001
TGFAb	74%	37%	17%	25%	38%	<0.0001
AREGa	3%	34%	43%	35%	41%	<0.0001
EGF	17%	40%	37%	36%	31%	0.23
CRYABa	70%	11%	33%	4%	48%	<0.0001
KRAS amplicona	68%	40%	24%	35%	0%	<0.0001
KRAS genec	32%	37%	33%	38%	21%	0.36
HRASd	32%	66%	17%	64%	7%	<0.0001
NRASa	70%	28%	17%	44%	21%	<0.0001
PIK3CA	30%	17%	36%	36%	41%	0.28
PIK3R1a	21%	14%	42%	25%	55%	0.0012
AKT1a	26%	63%	27%	40%	24%	<0.0001
AKT2*	26%	40%	27%	47%	38%	0.26
AKT3a	51%	14%	39%	9%	45%	<0.0001
MEK1	53%	46%	25%	29%	24%	0.023
MEK2e	42%	43%	25%	42%	24%	0.068
ERK1f	30%	26%	31%	42%	41%	0.49
ERK2g	40%	31%	26%	45%	31%	0.048

Samples were rank ordered into three equal groups and the percentage of each subtype in the highest expression group is reported for the NKI patient data set.

*Note: HER4 could not be assessed in UNC data due to too many missing values; HER3 was not present in the NKI data set; AKT2 was not present in the UNC data set

^a associations were also similarly significant in the UNC sample set

^b nominally significant in UNC data (p-value = 0.0046)

^c nominally significant association in the UNC data (p-value= 0.0051)

^d nominally significant in the UNC data (p-value = 0.003)

^e nominally significant in the UNC data (p-value = 0.0023)

^f significant in the UNC data (p-value = 0.0003)

^g significant in the UNC data (p-value = <0.0001)

Bonferroni corrected level of significance α = 0.0022