Table 5 PrimeKG can identify drug repurposing opportunities.

From: Building a knowledge graph to enable precision medicine

Drug

Disease

Shortest distance

Randomized distance (95% CI)

Adjusted P value

Ropeginterferon alfa-2b-njft

Acquired polycythemia vera

1

—

—

Tirzepatide

Type 2 diabetes mellitus

2

3.45 (3.40–3.49)

<0.01

Tezepelumab-ekko

Asthma

2

3.68 (3.63–3.73)

<0.01

Tapinarof

Psoriasis

2

3.98 (3.93–4.02)

<0.01

Faricimab-svoa

Macular degeneration

2

4.21 (4.17–4.26)

<0.01

Inclisiran

Familial hypercholesterolemia

2

4.61 (4.56–4.66)

<0.01

Maribavir

Cytomegalovirus infection

3

4.40 (4.36–4.45)

<0.01

Belzutifan

Von Hippel-Lindau

3

4.55 (4.50–4.59)

0.01

Ganaxolone

CDKL5 disorder

3

4.32 (4.27–4.37)

0.03

Pacritinib

Myelofibrosis

3

3.83 (3.78–3.89)

0.08

Tralokinumab-ldrm

Atopic dermatitis

3

3.69 (3.65–3.74)

0.19

  1. We retrieved 11 drugs with new indications approved by the FDA in the year after PrimeKG was assembled. We conducted a network proximity analysis between the repurposed drug and its indicated disease. Only 1 of 11 pairs already has an indication edge in PrimeKG, confirming that there is no temporal data leakage. We computer the shortest path distances between repurposed drug and (i). indicated disease; and (ii). a sample of 1000 non-indicated diseases. For the later, we have reported the mean randomized shortest path distance with 95% confidence intervals. We applied a non-parametric statistical analysis to assess how often the indicated shortest path distance is greater than the random shortest path distance and applied a Bonferroni correction to obtain the significance. At a threshold of P ≤ 0.05, relevant drugs are much closer to the indicated diseases than expected by random chance in 8 out of 10 cases. This analysis demonstrates the utility of PrimeKG for drug repurposing.
Back to article page