Drug repositioning based on bounded nuclear norm regularization

  • doi: 10.1093/bioinformatics/btz331

  • Objectives:

    • To identify new treatments by filling out the unknown entries in the drug–disease association matrix,
    • the drug–disease matrix to be completed is low-rank.

  • Results

    • A bounded nuclear norm regularization (BNNR) method to complete the drug–disease matrix under the low-rank assumption.
    • BNNR performs noisy matrix completion by incorporating nu- clear norm regularization, which effectively addresses overfitting and leads to better improved accuracy as shown in our results;
    • Proposed BNNR model incorporates a range constraint, which enforces all predicted matrix entry values within the specific interval;
    • Deals with Noise effectively
    • An efficient iterative scheme is designed to numerically solve the BNNR model.

  • IMP Points

    • 1’s in the drug–disease association matrix denote known drug–disease associations while 0’s represent the unknowns
    • Pasted image 20240926075323.png
      • Perhaps we could parallelize this matrix formation T = [Wrr, Wdr'; Wdr, Wdd]; this line in code does that.

  • FUTURE Scope

    • we plan to integrate drug–target information into the existing heteroge- neous networks to further improve the prediction ability of BNNR.

  • OBJECTIVE 1 : Study of different Algorithms in literature that support drug repositioning and it's need. (LITERATURE OBJECTIVE)

  • OBJECTIVE 2 : Parallelizing the existing BNNR algorithm from the base paper and putting the same parameters in results for comparison with same dataset and conditions/assumptions
  • Objective 3: Precisely proposing mathematical proof or model for new algorithm for better performance.
  • OBJECTIVE 4 : Comparison with existing algorithms and the proposed algorithm in objective 3.