Table 1.

Summary of implemented models’ performances. Models (*) were implemented for feature prestudying.

Models	Trained on [inputs]–[ground truth]	Fluorescence channel predictions
		MDA-MB-231			U2OS-AF	U2OS-CPS
		D/H	Vin	Fa	Fa	D/H	Mito	ER	NCR	GP-Fa
PhC-Fluo 1	[PhC]–[D/H(231)]	H
PhC-Fluo (*)	[PhC]–[Fa]			L
PhC-Fluo 2	[PhC]–[Vin]		L-A
Flou-Fluo 1	[Fa]–[Vin]		A-H
AF	[Fa]–[Fa(AF) or all channels CPS]				H	L-A	L-A	L-A	L-A	H
Flou-Fluo 2	[Gp-Fa+NCR]–[DP/H(CPS)]						H
Flou-Fluo 3	[Gp-Fa+NCR]–[Mito]							H
Flou-Fluo 4	[Gp-Fa+NCR]–[ER]								H
Flou-Fluo (*)	[Mito]–[D/H(CPS) or ER or GP-Fa]						A-H		A	L
Flou-Fluo (*)	[GP-Fa]–[D/H(CPS)]						L-A
Flou-Fluo (*)	[NCR]–[D/H(CPS)] or [Mito]						A	A

H, high; A, average; L, low performance; PhC, phase constrast; D/H, DAPI/Hoechst; Vin, vinculin; Fa, F-actin; Mito, mitochondria; ER, endoplasmic reticulum; NCR, nucleoli and cytoplasmic RNA; and GP-Fa, Golgi apparatus plasma membrane and F-actin.