Move Aside, Genome … It’s the Interactome’s Time to Shine
11/10/10 20:37 Filed in: GenomeWeb Daily Scan
Submitted by S. Pelech - Kinexus on Mon, 10/11/2010 - 20:37.
According the New Scientist's article, sequencing of 885 out of a planned 2500 people in the 1000 Genomes Project (Why don't they call this the 2500 Genomes Project?) have revealed 16 million DNA variations, and it appears that around 60 million human single nucleotide polymorphisms (SNPs) may exist. Unfortunately, the samples for the 1000 Genomes Project are mostly anonymous with no associated medical or phenotype data. Before the 1000 Genomes Project, over 8 million SNP were already known, but the physiological consequences of only a tiny fraction of these have been studied. At this rate, it will probably be several decades before most of the critical SNP's are actually identified and related to phenotype.
While elucidation of the "interactome" is a requisite for understanding phenotype, efforts on the proteomics front have been dismal. Firstly, over a decade was wasted on running 2-D gels and trying to identify abundant proteins by mass spectrometry. Very little useful information has really emerged from such efforts.
Secondly, so called "functional genomics" tracks the expression of genes by measurement of the levels of mRNA's for proteins. However, the correlation between mRNA and protein levels is extremely poor. It has become increasingly evident that micro-RNAs and many other factors strongly influence the translation of mRNAs into proteins. Consequently, we don't really know reliably what the expression levels of proteins are from gene microarray and quantitative PCR studies. Mass spectrometry is also not quantitative, unless strategies are adopted with reference peptides to examine specific proteins individually.
Thirdly, if a protein is produced, it does not mean that it is necessarily active. Most of the proteins encoded by the human genome are subject to phosphorylation and a plethora of other covalent modifications that influence functional activities, protein-interactions, subcellular localization and degradation. At Kinexus Bioinformatics Corporation, we believe that there are over 700,000 phosphorylation sites alone in human proteins. We will be revealing known and predicted phospho-sites in PhosphoNET (www.phosphonet.ca) later this year.
Fourthly, the vast majority of protein-protein interaction data that is available in repositories such as the Center for Cancer Systems Biology Interactome Database are primarily based on yeast dihybrid analysis and proteins from simple model organisms. In these studies, the bait proteins are unlikely to be properly post-translationally modified, and so the vast majority of protein-protein interactions are probably missed.
To really produce useful information that will permit a greater understanding of the cell and human pathology at the molecular level, it would be best to start with cells that are obtained from a vast set of around five to ten thousand diverse people where complete health and family histories are available along with an abundance of different cell types. This might be derived from fatal accident victims for health controls and from patients who have succumbed to common illnesses. The cell specimens should be subjected to full and standardized genomics and proteomics analyses so that direct comparisons can be generated. Such a task could only be achieved by the combined and coordinated efforts of several world governments through their medical research agencies.
Linking the genomics and proteomics data towards a rationale and predictable understanding of phenotype is a herculean effort that will probably take another hundred years or more. If this seems far fetched, consider that we have not been back to the moon in over 38 years, even though we have the technology.
Link to the original blog post.
According the New Scientist's article, sequencing of 885 out of a planned 2500 people in the 1000 Genomes Project (Why don't they call this the 2500 Genomes Project?) have revealed 16 million DNA variations, and it appears that around 60 million human single nucleotide polymorphisms (SNPs) may exist. Unfortunately, the samples for the 1000 Genomes Project are mostly anonymous with no associated medical or phenotype data. Before the 1000 Genomes Project, over 8 million SNP were already known, but the physiological consequences of only a tiny fraction of these have been studied. At this rate, it will probably be several decades before most of the critical SNP's are actually identified and related to phenotype.
While elucidation of the "interactome" is a requisite for understanding phenotype, efforts on the proteomics front have been dismal. Firstly, over a decade was wasted on running 2-D gels and trying to identify abundant proteins by mass spectrometry. Very little useful information has really emerged from such efforts.
Secondly, so called "functional genomics" tracks the expression of genes by measurement of the levels of mRNA's for proteins. However, the correlation between mRNA and protein levels is extremely poor. It has become increasingly evident that micro-RNAs and many other factors strongly influence the translation of mRNAs into proteins. Consequently, we don't really know reliably what the expression levels of proteins are from gene microarray and quantitative PCR studies. Mass spectrometry is also not quantitative, unless strategies are adopted with reference peptides to examine specific proteins individually.
Thirdly, if a protein is produced, it does not mean that it is necessarily active. Most of the proteins encoded by the human genome are subject to phosphorylation and a plethora of other covalent modifications that influence functional activities, protein-interactions, subcellular localization and degradation. At Kinexus Bioinformatics Corporation, we believe that there are over 700,000 phosphorylation sites alone in human proteins. We will be revealing known and predicted phospho-sites in PhosphoNET (www.phosphonet.ca) later this year.
Fourthly, the vast majority of protein-protein interaction data that is available in repositories such as the Center for Cancer Systems Biology Interactome Database are primarily based on yeast dihybrid analysis and proteins from simple model organisms. In these studies, the bait proteins are unlikely to be properly post-translationally modified, and so the vast majority of protein-protein interactions are probably missed.
To really produce useful information that will permit a greater understanding of the cell and human pathology at the molecular level, it would be best to start with cells that are obtained from a vast set of around five to ten thousand diverse people where complete health and family histories are available along with an abundance of different cell types. This might be derived from fatal accident victims for health controls and from patients who have succumbed to common illnesses. The cell specimens should be subjected to full and standardized genomics and proteomics analyses so that direct comparisons can be generated. Such a task could only be achieved by the combined and coordinated efforts of several world governments through their medical research agencies.
Linking the genomics and proteomics data towards a rationale and predictable understanding of phenotype is a herculean effort that will probably take another hundred years or more. If this seems far fetched, consider that we have not been back to the moon in over 38 years, even though we have the technology.
Link to the original blog post.