Shai Carmi Profile picture
Associate professor at the Hebrew University. Statistical, population, and medical genetics. Preimplantation genetic testing. Views my own. https://t.co/hlwDnVcr9O
Mar 16, 2023 7 tweets 3 min read
Fantastic investigative work by @mathiesoniain et al, providing strong evidence against claims
(nature.com/articles/s4158…) that specific immune gene variants were under very strong natural selection during the black death. It is insane that the original claims were selected to Science Magazine's 10 breakthroughs of 2022.
science.org/content/articl…
Nov 30, 2022 8 tweets 5 min read
Here are some popular articles covering our work on the DNA of 14th-century Ashkenazi Jews from Erfurt, Germany.

First, Andrew Curry @spoke32 in Science @NewsfromScience.
@ElieDolgin for the NY Times @nytimes.
Nov 30, 2022 22 tweets 6 min read
Our paper on the DNA of medieval Ashkenazi Jews is now published!

Congrats to all co-authors and thanks reviewers/editors!

We analyzed 33 genomes from a 14th-century Jewish cemetery in Erfurt, Germany.

See thread for key additions since the preprint.
cell.com/cell/fulltext/… But first, readers are welcome to glance at a series of Q&A I wrote in (hopefully) simple language.

The Q&A cover background on Ashkenazi genetics, explain the results, and discuss the implications, remaining open questions, and next steps.

docs.google.com/document/d/10g…
Jul 21, 2022 7 tweets 2 min read
פורסם לאחרונה הרצף המיטוכונדריאלי המלא של שערה שנמצאה על תאיר ראדה ז"ל. עמיתי @erikl861 בדק לעומק את תוצאות הריצוף.
הרצף המלא מתאים להפלוגרופ בשם I. זה כבר היה ידוע. מה שמעניין הוא שלהפלוגרופ I יש מספר "תתי-הפלוגרופים", אך הרצף שפורסם לא מתאים לאף אחד מהם. מה זה אומר? המשמעות היא שאם מקודם נמצאו במאגר מסוים X נשאים של הפלוגרופ I, כולם היו יכולים להיות הבעלים של אותה שערה. כעת אנחנו יודעים שאם חלק מהם היו נשאים של הפלוגרופים כמו I1 או I4 וכו', הם לא בעלי השערה. זה מצמצם את מספר האנשים שיכולים להיות קשורים לרצח.
Jul 21, 2022 21 tweets 4 min read
Following up, I think the discussion on embryo screening should move on to become more nuanced. I see the same arguments over and over, ignoring a growing body of literature.
Let me try to summarize where the field is standing in my subjective view. Here are a few concerns that are in principle correct, but their impact is expected to be small, and we should probably stop bothering.

1) Computing a PRS from an embryo biopsy will add errors. (It will, but the impact is small, at least with high coverage).
Jul 20, 2022 11 tweets 3 min read
A new opinion piece in Human Reproduction arguing against screening embryos with polygenic risk scores.

The article mentions a few important problems, although I think not much is new. And I disagree with some of their other arguments.

academic.oup.com/humrep/advance… But first, a huge surprise to see Julian Savulescu in the list of authors. Savulescu has been arguing *for* embryo screening for almost 20 years. onlinelibrary.wiley.com/doi/abs/10.111…
Including a talk he gave in ESHRE just last year. Wondering what happened.
May 27, 2022 6 tweets 2 min read
This is a brilliant preprint. It describes an experimental method for separately sequencing the paternal and maternal genomes. This goes beyond the standard "experimental phasing" - the method can also tell which chr is maternal and which is paternal.

biorxiv.org/content/10.110… The experimental phasing is based on long reads from nanopore sequencing, plus a very cool method called "strand-seq". I didn't know about it before, and found a good explanation in the introduction here: academic.oup.com/bioinformatics… Image
May 20, 2022 6 tweets 3 min read
Interesting work, reporting three new genomes from Neolithic Malta.
The paper applied new haplotype-based methods, facilitated by imputation of both shotgun and SNP capture genomes.
1/ First was a ROH (runs-of-homozygosity) analysis.
One Maltese individual (Xaghra9) had 700cM in ROH!
Interestingly, the ROH segments were shorter than expected under first-degree consanguinity, suggesting very small eff population size (estimated using ROH as ~500).
2/
May 18, 2022 32 tweets 12 min read
Glad to share a new preprint!

We report genome-wide data for 33 Ashkenazi Jews from 14th-century Erfurt, Germany, along with insights on the Ashkenazi founder event and medieval population structure.

Let's dive into the story of the Erfurt Jews.

biorxiv.org/content/10.110… Image Our story begins in the 11th century, when Jews first settled in Erfurt. Their community was successful; the synagogue they built (photo) is the oldest one still standing in Europe.

This was interrupted by a pogrom in 1349, when nearly the entire community was murdered. Image
Aug 1, 2021 15 tweets 5 min read
This is a new perspective paper by @mathiesoniain, suggesting that the low transferability of polygenic scores across populations is an inevitable consequence of the "omnigenic" model for the genetic architecture of traits and diseases.

1/

cell.com/ajhg/fulltext/… Why are polygenic scores derived in one population (usually European) not working well in other populations?

Some of the reduction is accuracy is due to differences in LD and allele frequencies. But this still does not explain everything.

2/
May 3, 2021 6 tweets 4 min read
I hate to say this, but this paper in @genomeresearch has many problems.

They developed a method to select an optimal subset of individuals that together cover the most variants.
However, they ignored a huge body of pre-existing literature.

1/6

genome.cshlp.org/content/31/5/9… Some examples:
genetics.org/content/190/2/…
onlinelibrary.wiley.com/doi/abs/10.100…
nature.com/articles/ejhg2…
genetics.org/content/195/2/…
cell.com/ajhg/fulltext/…
gsejournal.biomedcentral.com/articles/10.11…
academic.oup.com/bioinformatics…
biorxiv.org/content/10.110…

2/6
Apr 19, 2021 14 tweets 5 min read
An interesting new preprint on how to combine PRS and family history to improve disease risk prediction. By Alkes Price and colleagues.

Bottom line: adding disease status of family members as indicators to logistic regression works well.

Thread ->

1/14

biorxiv.org/content/10.110… Polygenic scores quantify the risk due to known common variants.

Family history (FH) represents overlapping, yet distinct information: e.g., an affected parent may transmit to the child risk variants not in the score.

So both PRS and FH can improve prediction accuracy.

2/14
Apr 6, 2021 10 tweets 4 min read
Remember when the Broad Institute discovered polygenic scores? Now it seems as if they invented quantitative genetics.

See below for a thread. (Happy if someone could send me the full text.)

1/7

ahajournals.org/doi/abs/10.116… The most remarkable "discovery" is this: the correlation between the (CAD) polygenic scores of siblings is 0.5. And that between second-degree relatives is 0.25.

As much as I know, this result is over 100 years old.
genetics.org/content/211/4/…

2/7
Apr 5, 2021 4 tweets 3 min read
Nice work by @benmpeter's lab.

Main results:
(1): A model for the segment length distribution under admixture that was ongoing for multiple generations.
(2): Extensive simulations for testing inference accuracy for admixture time + duration, ...

1/4

biorxiv.org/content/10.110… including when introducing errors in the recombination map. It seems that long admixture introduces high variability in the inference of the admixture duration, in particular when the recombination rate is mis-specified.

2/4
Apr 5, 2021 9 tweets 2 min read
Here are impressions from an *in-person* class I just taught - my first since Jan 2020.

The setting: statistics for medical students. Total enrolled: 220; total in class: ~120; watching from home: ~60.

1/8
(1) Strangely, while the students were first year undergrads, they already knew each other very well (not sure where from). It didn't feel at all like I was teaching students on their very first day on campus.

2/8
Jan 8, 2021 10 tweets 3 min read
What can we learn from sequencing (100% genetically identical?) monozygotic twin pairs?

Turns out, a lot, particularly on early embryo development. Here, deCODE deeply sequenced ~400 twin pairs, along with their children/parents when available.

1/10

nature.com/articles/s4158… The authors found variability in the no. of postzygotic mutations (not in the twin's parents), e.g. 39 twin pairs differed by >100 mutations, 38 pairs did not differ at all.
The number of mutations increased with age, indicating that most of them accumulate through life.

2/10 ImageImage
May 6, 2019 11 tweets 4 min read
Just posted our manuscript "Screening human embryos for polygenic traits has limited utility"

We provide an empirical foundation to the ethical debate regarding the generation of “designer babies” by screening IVF embryos. 1/11

@ehudkar @ToddLencz

biorxiv.org/content/10.110… Large-scale GWASs yield increasingly accurate polygenic scores (PS), and it is now feasible to calculate such scores from genome-wide data of IVF embryos.

One company is already offering embryo screening for disease risk scores tinyurl.com/yygzdw7q 2/11