Presentation

Functional Annotation Genomic and Proteomic Data

Exploring pathways to better understand uncharacterized orphan proteins in order to better investigate causes of tumour cells and diseases

Team 7: Max Sun, Ali Fathi, Andrew Trang, Aurora Tung, Alexander Brechalov

Motivation

First step to treat diseases: understand protein pathways

Uncharacterized cancer proteins: Lung, Breast, Pancreatic

John McLeod: diabetes -> insulin

Methodology

Reseach on orphaned/unannotated proteins

Protein Name -> GO -> diseases pathways

Protein Name -> Uniprot -> Sequence

Sequence -> InterPro -> ID, family & domains

BLAST -> sequence similarity

Project Layout

graph TD;
    Input_protein_seqs-->APIs;
    APIs-->InterPro;
    APIs-->UniProt;
    APIs-->Ensemble;
    InterPro-->query_result;
    Ensemble-->query_result;
    UniProt-->query_result;
    query_result-->GO_terms;

Loading

Model

Dataset: Human proteome ~20k of reviewed proteins (UniProt)

Model: multiple Logistic Regressions (PyTorch)

Input: 15k of InterPro domains (0/1 vector)

Output: 20k of GO terms

Results

Training the model

Example:

Future work

Improve model

Reproducible: dependencies & containerization

Current: Sequence

Future: Mutation cause & 3D structure

Rebuttal

Accuracy: check against existing data for partially characterized proteins

Infer functions from protein family and similar protein

Some proteins hard to access

Notes

Some proteins are not annotated nor characterized:

modify/translate other proteins

evolutionary divergence

Sickle Cell Anemia: hemoglobin

Tumour: p53

1. Protein Name -> GO (Gene Ontology) -> Diseases Pathways:

   • Step: Start with the protein name and search for its Gene Ontology (GO) annotations.
   • Tools: Use resources like the [Gene Ontology Consortium](https://www.geneontology.org/) or [UniProt](https://www.uniprot.org/help/gene_ontology) to find GO terms related to the protein.
   • Outcome: Identify the biological processes, molecular functions, and cellular components associated with the protein. This can help link the protein to specific disease pathways⁹.

2. Protein Name -> UniProt -> Sequence:

   • Step: Use the protein name to find its sequence in the UniProt database.
   • Tools: Search UniProt using the protein name.
   • Outcome: Obtain the protein sequence and other relevant information such as functional annotations and known interactions¹⁷.

3. Sequence -> InterPro -> ID, Family & Domains:

   • Step: Input the protein sequence into InterPro to identify protein families, domains, and important sites.
   • Tools: Use InterPro for functional analysis and classification of proteins.
   • Outcome: Get detailed information about the protein's family, domains, and functional sites⁴⁵.

4. BLAST -> Sequence Similarity:

   • Step: Perform a BLAST search to find sequences similar to your protein.
   • Tools: Use the NCBI BLAST tool.
   • Outcome: Identify homologous sequences and infer functional and evolutionary relationships¹³¹⁴.