State of the art of hypothetical protein prediction in prokaryotes

October 21, 2018 | 2 Minute Read

Original Paper:

Tackling Hypotheticals in Helminth Genomes.

• Type: review article
• Organism: helmith

This article is about helmith. But what I care are prokaryotes. So I decide to make a note to record all aspects mentioned in this paper, but targeting prokaryotes.

An silico approach based on gene expression data and GO term: gamma-AM

Original Paper

• Example: There are two sets of genes. Gene set A has both expression data and GO term. Set B has only expression data.
• Step 1: calculate the similarity between all genes within gene set A. The similarity is defined as “gamma distance”.
  • gamma distance = (gamma)(distance between go terms) + (1-gamma)(distance between expression)
  • the gamma value decides the weighting (importance) of GO term.
• Step 2: cluster gene set A based on gamma distance
• Step 3: assign genes in set B into clusters that are formed by set A genes.
  • set B genes are assigned to the cluster if the centroid (with only expression data) is the closest.
• Step 4: infer GO term of the set B genes by referring to other set A members in the same cluster (cluster enrichment analysis)
• Note: this algorithm is not dedicated to helmith only. Applicable to bacteria.

Methods Based On Domain, or the sequence homology

• homology: KEGG, Uniprot, BRENDA
• based on domain and sites: InterProScan
• The above ones have higher false negative due to poor convergence of sequence outside of the active site.

Methods to improve false negative of homology models

• PRIAM
  • target: ** enzymes**
  • method:
    • group all available enzymes according to their EC number. (An EC number corresponds to one kind of activity. For example: alcohol dehydrogenase)
    • find the longest common sequence within a single EC group. This is called a module.
• EFICAZ and DETECT are also for enzymes only

Method based on phylogenomics

Other

• HMM
• 3D homology

Methods only suitable for core genes (conserved hypothetical proteins)

• Pathway reconstruction and pathway inference
  • input: genome
  • output: pathways and genes involved
  • missed genes in the pathway can be the possible role of the hypothetical protein
• orthology based inference

Validation of inference

• cDNA search in database
• RNA-Seq
  • Improved technology allow us to see the 5’ end of RNA, enabling us to find the promoter and possibly find the whole operon
• Proteomics: LC MS/MS
• Functional genomics tools:
  • CRISPR
  • RNAi

Hypothetical genes can be RNA?

• Do prokaryotes have ncRNA? YES
• What do they do? complement to mRNA and kill them. Associated with virulence, quorum sensing
• Tools to predict small RNA in prokaryotes? Here