Table 1.

TimeSignature predictive genes

Gene	Freq.	Gene	Freq.	Gene	Freq.
DDIT4	1.00	GZMB	0.58	CAMKK1	0.17
GHRL	1.00	CLEC10 A	0.50	DTYMK	0.17
PER1	1.00	PDK1	0.50	NPEPL1	0.08
EPHX2	0.92	GPCPD1	0.50	MS4A3	0.08
GNG2	0.83	MUM1	0.33	IL13RA1	0.08
IL1B	0.83	STIP1	0.33	ID3	0.08
DHRS13	0.83	CHSY1	0.25	MEGF6	0.08
NR1D1	0.75	AK5	0.25	TCN1	0.08
ZNF438	0.75	CYB561	0.25	NSUN3	0.08
NR1D2	0.75	SLPI	0.25	POLH	0.08
CD38	0.75	PARP2	0.25	SYT11	0.08
TIAM2	0.75	PGPEP1	0.17	SH2D1B	0.08
CD1C	0.75	C12orf75	0.17	REM2	0.08
LLGL2	0.58	FKBP4	0.17

A set of 41 genes is sufficient for accurate TimeSignature prediction. As described in the text, the predictors may vary slightly depending on the training data; the frequency with which each gene was selected as a predictor across 12 repeated runs using different training samples is given. Eighteen are selected the majority of the time; the remaining “auxiliary” predictors vary from run to run. Genes are sorted in order of selection frequency.