each individual patient could be deemed as a matrix complex
the individual matrix (say it diagnosis) could be indicated or determined by some “endogenous” matrixes or vectors (i.e.radiomics, slides, mRNA, protein, methylation, TMB or non-coding RNAs)
the individual matrix could also be changed or modified by certain interventions such as diet, exercises, medicines, radiotherapy or surgery, all of which could also be deemed as “external” matrixes or vectors
- radiomics(lowest power)/->image fingerprint::???
- slides(lower power/high power)/->image fingerprint::pHASH
- protein(higher power)/->protein fingerprint::bunch functionally related genes of of differential expression
- mRNA(highest power)/->mRNA fingerprint::bunch functionally related genes of of differential expression
- methylation(->mRNA quantity)/->methylome fingerprint::bunch of differentially identified sites
- SNPs&TMB(->mRNA quantity&quality)/->variants situated in functionally related genes
- alternative splicing(->mRNA quantity&quality)
- non-coding RNA(->mRNA quantity)
- CSCs(phentype of cancer)/
- metastasis(phentype of cancer)/
- immunity(phenotype of immunology)/
similarity search by calculating the distance between images is a possible way for the purpose of pathological diagnosis, while simmply doing so may not help a lot so far. an alternative strategy might be something that we name it as “divide and rule”: on one side, specify the features of whole image (texture, outlier etc) to compare in between to recognize the overal structure of the image; and on the other side, find the ROIs or FTUs, such as tubules and granomuli in a slide of kidney, or nuclei of a cancer. actually this is the way how a pathologist recognize and diagnose a slide: zoom out+zoom in
basically, use pHASH, VSM based smilarity(small scale) and simHASH(large scale)
steps towards
- mRNA is the pivot of the whole sturcture
- cancer mRNA fingerprint (vs. normal)
- image fingerprint (vs. normal???)
- correlation analysis
- PCA/UMAP/TSNE/LDA->dimension reduction
- similarity analysis by simHASH or VSM(cancer image fingerprint vs. mRNA fingerprint)
- image classification by mRNA using deep learning
the mRNA fingerprint does not mean the whole mRNA profiling, might be just some of them, as those are oncogenes or tumorsupressors, or those differentially expressed
the slide image fingerprnt does not mean it has to be the whole image, might be just some critical image tiles or the textures, outlier or mitotic features of nuclei
CARDIOGENE 