Overall association score with scored weights of 0

General
1

Hi community!

I am diving into the overall association score and playing with the weights. When I set all the weights to 0, except 1 source, the score of that source does not match the overall association score. The picture below shows an example of cystic fibrosis where:

  • only REACTOME has a weight of 1
  • CFTR - REACTOME score = 0.93
  • CFTR - overall score = 0.56

image
image2836×699 76 KB

I am aware how the overall score is calculated; and I understand the technical reason why the overall score is lower than the individual score (i.e. 0.93 / 1.644 = 0.56)
However, I am curious to understand the reason why the harmonic sum (1.644) is not affected by the weights. e.g. if a weight = 0, that it’s not included to calculate the harmonic sum.

For me it does not look intuitive that the overall score does not match the REACTOME score. I can imagine a use case where you want to exclude some source types like literature or somatic mutations ( since cystic fibrosis is not a cancer type).

I would love to hear your opinion on this topic!

Kind regards,

Loes van den Biggelaar

2

Hi Loes,
welcome to our community and thanks for your question!
So, as mentioned in our documentation, the harmonic sum of the datasource scores is computed assuming infinite vector, of scores yielding a constant value for scaling - and this value is not dynamically updated with the weights.
We are currently revising our methods and we will think about whether it would make sense to change this.
Please do let me know if you have any additional questions or comments for us!

Best wishes,

Annalisa