Table 1.

Training set gene list

MSMGs	Non-MSMGs
Bmp4, Cdyl, Cdyl2, Dmrt1, Dppa4,	Afp, Arid3a, Arid3b, Ascl1, Ascl2,
Dppa5a, Esrrb, Etv4, Etv5, Fgf4, Foxd3,	Bat1a, Bmp2, Bmp5, Bmper, Ccnd2,
Foxh1, Gbx2, Grhl2, Jarid2, Klf2, Klf5,	Cdh2, Cebpa, Cited1, Dach1, Dlx1,
Lefty2, Lin28, Mkrn1, Mycn, Nanog,	Dlx4, Dlx6, Ednra, En1, Eomes, Ets2,
Nodal, Nr0b1, Nr5a2, Phc1, Phf17,	Eya2, Fgf5, Foxb1, Gata1, Gata3,
Pou4f2, Pou5f1, Rif1, Sall1, Sall4, Sgk1,	Gata4, Gata5, Gata6, Gfap, Gli3, Gsc,
Slc27a2, Socs3, Sox2, Spp1, Tcf15,	Hand1, Hand2, Insm1, Isl1, Lbx1,
Tcfap2c, Tcfcp2l1, Tcl1, Tle4, Trp53,	Lhx2, Lhx5, Lmx1a, Mbd2, Meis1,
Utf1, Zfp296, Zfp42	Mixl1, Myf5, Neurog1, Nfia, Npas3,
	Nr2f1, Nr2f2, Nrp1, Nrp2, Olig3, Otp,
	Otx1, Pax3, Pdx1, Peg3, Phox2b,
	Prl3d1, Prox1, Rybp, Shh, Sox1,
	Sox18, Sox3, Sox5, Sox9, Stra13, Syp,
	Tcf4

List of genes used as training set include 46 positive examples labelled as MSMG class and 70 negative examples labelled as non-MSMG class. These genes are derived from expert curation.