Clarification on Drug-Effects Edge Count Discrepancy in PrimeKG

[gbatsis]

First of all, congratulations on your exceptional work with both PrimeKG and TxGNN! The contributions these tools make to the field are truly impressive.

---

However, I noticed a discrepancy regarding the number of edges between drugs and effects in PrimeKG. In your paper, it is mentioned that there are over 200k edges in this category. Yet, during my exploratory data analysis (EDA) on PrimeKG, I found the count to be 64,784.

Could you kindly clarify this difference? Is there additional preprocessing or filtering applied in the paper that might explain the variance? Any insights into this discrepancy would be greatly appreciated.