Using AlphaGenome to predict the cooperativity of multiple SNPs?

Community
1

Hello,

Thanks for creating such a powerful tool and making it readily available for the research community. I am an oncologist as well as a research scientist. I have been testing AlphaGenome for the past few days and was impressed by its ability to accurately replicate results for single SNP changes. However, many cancers are associated with haplotypes where multiple SNPs, each tens of thousand BPs apart, jointly alter the DNA expression level/TF binding/splicing.

Question: is alphagenome able to predict the cooperativity of multiple SNPs at the same time?

Thanks

2

I have an idea, maybe try it?

Since there’s no restriction on variant length, you can try input a variant like this:
Chromosome: chr1 (or others)
Position: 12345…
ref: ACTGCGATTACG (can be longer)
alt: ACTACGATTACG

This case showed a two-SNP situation, and you can add more as you wish.

There are two problems: 1. You have to find the sequence on the reference genome in this region. Use samtools faidx hg38.fa chr1:123-456 can do this.
2. The cover region of this variant can be so long, I’m not sure if the result will be weird. And the visualization can be hard too.

In my test cases, I tried a 50kb deletion, and still got a result. So it will work, but the result is not known. You can share the result with us!

3

Hi Jingyu,

Thanks for replying. I also thought about this but was not sure if the input sequence would be too long. Good to know that 50kb deletion ran just fine. I will give it a try.

4

Hi Jingyu,

I tried a 7kb sequence with 4 SNPs but it broke AlphaGenome. How was your experience with the large deletion?

5

Can you show me your error? I’m not sure what happened.

In my dataset, there are some large indels. They all work well.

6

What is your input size? Mine was 7kb. Not sure if that’s too big.

7

never mind, i figured it out