Skip to main content
Fig. 7 | BMC Bioinformatics

Fig. 7

From: MIC_Locator: a novel image-based protein subcellular location multi-label prediction model based on multi-scale monogenic signal representation and intensity encoding strategy

Fig. 7

The flowchart of proposed MIC_Locator. The IHC image is selected from gene “ENSG00000013364”. The corresponding number of IHC image is “6980_A_4_6”, and it belongs to the “Cytosol” subcellular location. In the preprocess stage, the DNA and protein channel of protein are separated. On the one hand, the DNA and protein channel are used to extract the 840-dimension SLFs feature. On the other hand, the protein channel is transformed into the frequency domain by the Fourier transform. The frequency information of protein is multiplied by the Riesz transform, generating two frequency responses in orthogonal directions. The frequency information of protein and two frequency response parts of Riesz transform are multiplied by the Log-Gabor filter with multi-scale frequency factor. Afterwards, the protein information and two frequency response parts are transformed into the spatial domain, which commonly consist of the monogenic signal of protein. The APO components of image monogenic signal are calculated. The 8-bits LBP code extracts the statistic information of APO component, and the 2-bits image intensity code is calculated from the two imaginary parts of monogenic signal by the formula (19). The LBP, image intensity and SLFs are united as the final 1864-dimension sample feature, feeding into the CC. The top and threshold criteria are applied to judge the subcellular localizations of test sample

Back to article page

BMC Bioinformatics

ISSN: 1471-2105

Contact us