Appendix 1—table 7. Sequence similarity comparison.

Frequencies of dipeptides (pairs of neighboring amino residues) were computed for phase-separating proteins and the human proteome, and enrichment was measured by the percentage of human proteins with lower frequency than found in a given sequence. The fifteen dipeptides enriched (≥99%) in the most sequences within the phase separation test sets are shown in the table vs. enrichment values obtained for the phase separation training set and three experimentally verified proteins. Values in the top fifth percentile are shown in bold.

Protein Name	Dipeptide enrichment (Percentage of human proteome with lower frequency)
	GV	VG	VP	PG	FG	RG	GR	GG	YG	GS	SG	GA	GF	GD	DS
Training Set Proteins
Elastin	100	100	100	100	97	31	32	99	99	20	20	100	89	38	30
Nsp1	30	34	31	26	100	31	30	75	52	90	38	99	60	66	68
TIA1	73	75	46	26	86	31	86	77	99	29	53	26	84	54	30
LAF1	30	78	65	29	67	99	99	100	77	88	97	65	78	97	32
EIF4H	30	65	31	52	98	99	95	99	52	99	42	79	99	98	89
Ddx3x	51	70	34	43	89	98	97	96	93	93	95	68	96	59	78
hnRNPA1	30	55	31	44	100	99	99	100	99	99	98	44	99	60	79
DDX4	33	77	48	53	98	96	91	89	59	87	96	29	98	96	45
FUS	30	31	31	83	78	100	99	100	100	98	99	33	93	91	57
EWS	52	31	35	97	51	100	99	100	100	72	61	30	97	91	48
TAF15	36	38	31	30	53	100	99	100	100	92	99	26	71	100	94
Experimentally Verified Proteins
FMR1	69	89	93	44	43	96	94	83	62	62	34	70	48	41	67
SCAF pAP	75	50	89	91	43	49	73	97	75	92	96	96	40	36	44
Engrailed-2	30	31	31	97	70	78	90	100	52	99	91	99	40	95	97