Florian Privé

I guess I'll be here in the future: bsky.app/profile/privef…

This is my farewell post. I am not deleting this account, as I don’t want to lose the archive, and I don’t want my name taken by imposters. But, for as long as Musk owns the platform, I’ll no longer use it. I’m now posting at georgemonbiot.bsky.social. Since Elon Musk bought this site, he has transformed it from one on which millions could converse as equals into his personal megaphone. Now the world’s richest man uses it to wage his class war, transferring blame for the ills of capitalism onto vulnerable minorities, boosting the grimmest and most antisocial accounts while suppressing the humane voices with whom he disagrees. He has used X not only to subvert the election for the US presidency, but to create a role for himself in US politics which, though he has never stood for election and would not be eligible, grants him immense power over the citizens of that nation. This site has been used as a tool to help replace democracy with oligarchy. Musk claimed, on purchasing the platform, to be a “free speech absolutist”. He is in fact a prolific censor, suspending dissenting accounts and using lawsuits to shut down free speech, while employing an algorithm to increase the reach of his own posts 1,000-fold. He cannot abide a level playing field. He has permitted or encouraged the growth of a ghost army of bots and trolls to degrade the experience of all who disagree with him. Their task is not just to abuse opponents but, perhaps more importantly, to drown intelligent conversation with a tsunami of stupidity. They respond to data with denial, expertise with execration, insight with insult, humour with hatred and thinking with thuggery. They suck the meaning out of everything. The stupidity trends in one direction: in the service of economic power. It is a weapon used to suppress the possibility of a better world. Many years ago, when I lived in Brazil, I learnt that the military dictatorship had actively sought to suppress educational levels in schools, as it saw a well-informed population as a political threat. A confused, distracted and ignorant population is easy to manipulate. Now this counter-educational model is being rolled out worldwide, and Elon Musk is its primary sponsor. I stayed here long after I first considered leaving, as I believed it was wrong to cede the ground to an oligarch and his minions. But I came to see that those of us who do not subscribe to Musk’s grim project are being used as groundbait: stimulating the feeding frenzy of 15-minute, 24-hour hate that now powers this platform. We can no longer build anything of value here. Brute force – the unmediated power of money – has beaten humanity, intelligence, humour and democracy. On Bluesky I can be the person I want to be, a better person than I am on X, where it is almost impossible not to get dragged into the mud. I feel I can be understood because I am not confronting a deliberate effort to misunderstand me. So I can speak more quietly. There is no guarantee that Bluesky will not also one day be monetised by its owners or captured by self-serving billionaires. But for now it remains a place in which interesting and enjoyable conversations can be had, kindness can be shown to strangers and a better world can be imagined. I hope to see you there.

Our paper on misleading biases in AD GWAS-by-proxy is published @NatureGenet. We identify the source of biases and explore strategies to reduce them @yy_stat @SunZhongxuan Paper📰: doi.org/10.1038/s41588… Software🧑‍💻: github.com/qlu-lab/GSUB Sumstats⬇️: qlu-lab.org/data.html

@mblum_g @felbalazard

Question for #epitwitter. Is it sound to study the effect of pollution on health outcomes using Mendelian Randomisation? My answer would be no since genetic changes can not modify the exposure. But there are many papers published using MR for this question @felbalazard

@pp0196 No hyper-parameter I think. I just need to find a good way to choose the number of groups.

@privefl so there should be sensitivity to the hyper-parameters (if there are) of the automatic grouping algorithm

Automatic detection of ancestry groups based on PCA. How useful do you think this is?

@pp0196 The goal would be to get ancestry proportions from automatic ancestry groups, based on PCA.

@privefl is it going to be a pca based classifier like in the your ancestry paper ?

@pp0196 I guess PCA is not much sensitive to those, so it should be fine.

@privefl Yes sensitivity to SNP and sample distributions

@pp0196 It is deterministic, if that's your question. If you meant whether I get the same results when e.g. subsetting, I'll need to test on some larger dataset.

@privefl how stable is it ?

The UK Biobank (UKB) has been an amazing field changing resource for the past decade. Unfortunately, and very suddenly, they have decided that to work with their data you now *must* use their “Research Analysis Platform” (RAP). @uk_biobank I guess there have been general warning

New Oxford-Aarhus-Copenhagen collaboration - not yet launched - 2 postdocs, 2 senior postdocs and postdoc coordinator in Oxford. Jobs 173031-3 at jobs.ox.ac.uk Methods-focussed in group Holmes, Visscher, Evans, Nicholson, @cwcyau, @ten_photo etc, Denmark ads soon

@mblum_g . @antuki13 devrait savoir

Quelqu'un sait comment faire une carte de France avec les circonscriptions avec #rstats?

@Rbn_Hfmstr I guess, if you have this, you can also report the r2(G_bgen, G_wgs) for each variant in the bgen data? This is really what I am looking for.

@privefl If it helps: in this figure, the aim was to assess ukb axiom array imputation accuracy from different reference panel by comparing imputed genotypes to WGS data.

Has anyone checked the consistency between the imputed dosage (BGEN) data from the UK Biobank vs the new sequencing data?

@AlexTISYoung Then we should talk :-)

@privefl I'm working on something like this.

@elonmuskADO No. Twitter.

@ShaiCarmi Or by inferring the ancestry proportions of each individual using doi.org/10.1093/bioinf… (where AJ is one of the 21 ancestry groups for which proportions are reported).

The AJ cluster could be identified easily by highlighting the UMAP positions of carriers of AJ-unique variants, such as BRCA1 del185AG gnomad.broadinstitute.org/variant/17-431… or CFTR W1282X gnomad.broadinstitute.org/variant/7-1176…

There's one point that I think was overlooked in the debate over the UMAP plot: Where is the Ashkenazi Jewish cluster? Given 125k people of "white" ancestry, there should have been at least 2000 Ashkenazi Jews. Do they form the circled blob?

I visited them a few weeks ago; great team!

Just registered ✅

"Benchmarking of local genetic correlation estimation methods using summary statistics from genome-wide association studies" -- doi.org/10.1093/bib/bb… Very nice *external* comparison of 3 sumstats methods for inferring local genetic correlations. - ρ-hess, SUPERNOVA and LAVA -