-
Notice: S4ME will have a short interruption of service on May 30. More information here.
Running FLAMES on DecodeME data
- Thread starter ChronicallyOverIt
- Start date
ME/CFS Science Blog
Senior Member (Voting Rights)
I got slightly different results using the UK biobank LD panel (UKBrelease2b10kEuropean). I specified the full rather than effective sample size (n = 275488).
FLAMES highlighted 31 instead of 32 causal genes (and VRK2 appears twice again, so it's 30 genes). The main differences are:
No longer in FLAMES with UK biobank LD FLAMES
RPP40
PLCL1
NEURL1
NEK1
New in FLAMES with UK biobank LD FLAMES
MLLT10
HTT
SSR1
It probably makes more sense to focus on the UK biobank LD panel so would focus on this list. I'll paste it below again:
ARFGEF2
CA10
UNC13C
SHISA6
SOX6
MMS22L
OLFM4
PEBP1
ZNF644
LRRC7
DCC
MLLT10
HTT
CACNA1E
VRK2
ALK
VRK2
MICALL2
KIAA1239
STT3B
VPS54
RIMS1
PTPRE
NR2F1
PTBP2
RP11-147C23.1
ADARB2
SMCHD1
SSR1
LAMA2
HABP2
Nice, thanks.
When doing the enrichment again with those, only one gene set is highlighted: synapse.
Another gene set, which I guess is targets of the microRNA mir-3620, was also significant, but was not highlighted, probably due to similarity to the synapse gene set.
| term_id | term_name | highlighted | adjusted_p_value | term_size | query_size | intersection_size | effective_domain_size | intersections |
|---|---|---|---|---|---|---|---|---|
| GO:0045202 | synapse | TRUE | 0.0159 | 1608 | 27 | 9 | 22155 | ARFGEF2,UNC13C,SHISA6,DCC,HTT,CACNA1E,VPS54,RIMS1,LAMA2 |
| MIRNA:hsa-mir-3620 | hsa-mir-3620 | FALSE | 0.0240 | 840 | 27 | 8 | 16638 | ARFGEF2,SHISA6,PEBP1,MLLT10,CACNA1E,PTBP2,SMCHD1,SSR1 |
It seems the other synapse-related gene sets are not significant anymore mainly because they include the now removed PLCL1 and/or NEURL1.
Edit: I might have been wrong about the reason mir-3620 isn't highlighted. I think the "highlighting" is just for GO terms. It might be worth looking at this miRNA.
Last edited:
@forestglip, I have a question about this data.
when I click on GO:0045202, the link takes to another website and when I click on the link there to all direct and indirect annotations to synapse (excluding "regulates"), I get this:
This page says that the synapse data is from domestic cattle and domestic cats. Do we have any idea of the relevance of this information to human synapses?
when I click on GO:0045202, the link takes to another website and when I click on the link there to all direct and indirect annotations to synapse (excluding "regulates"), I get this:
This page says that the synapse data is from domestic cattle and domestic cats. Do we have any idea of the relevance of this information to human synapses?
Good question. I don't know a lot about the gene ontology database. [Edit: For all I know, the annotations for all the different species are just the different forms of the same genes, just for a different animal. But I don't know how true that is.]
In any case, it looks like all of the DecodeME genes that showed up in the synapse gene set, and thus made it significant, are annotated under "Homo sapiens" so probably relevant to humans.
The gene set is also available on msigdb, where it says source species "Homo sapiens". If expanding the link to Show Members near the bottom, all of the following synapse DecodeME genes are there:
Last edited:
Months ago I made an attempt at fine mapping the DecodeME data using SusieR with the LD matrices of UK Biobank, and GTEx v10 for gene mapping.
My list was: ABT1, ANKRD45, ARFGEF2, BTN2A2, CSE1L, DARS2, KLHL20, PRDX6, RABGAP1L, RC3H1, SERPINC1, SLC9C2, STAU1, TNFSF4, TRIM38, ZBTB37, ZNFX1, OLFM4.
The pipeline is here: https://github.com/paolomaccallini-hub/DecodeME
My list was: ABT1, ANKRD45, ARFGEF2, BTN2A2, CSE1L, DARS2, KLHL20, PRDX6, RABGAP1L, RC3H1, SERPINC1, SLC9C2, STAU1, TNFSF4, TRIM38, ZBTB37, ZNFX1, OLFM4.
The pipeline is here: https://github.com/paolomaccallini-hub/DecodeME
ME/CFS Science Blog
Senior Member (Voting Rights)
Thanks, quite a lot of differences unfortunately. Did it only focus on the hits above 5*10^-8?
Yes, exactly. I am not familiar with FLAMES. I used a classic approach (SusieR selects the true signal by posterior inclusion probability, PIP), no machine learning. Also, I used eQTLs from GTEx V10 to map true causal variants to genes. I wonder if this affected the results.