Sign in

Shai Carmi

@shaicarmi.bsky.social
934 followers 225 following 481 posts

Associate professor at the Hebrew University of Jerusalem. Statistical, population, and medical genetics; preimplantation genetic testing. Views my own. scarmilab.org

PostsRepliesMedia
Shai Carmi @shaicarmi.bsky.social · 12/10/2026
Amazing story!
110
Reposted by Shai Carmi
Evan Irving-Pease @evanirvingpease.bsky.social · 06/10/2026
New #aDNA paper! 💀🧬 Ancient DNA from colonial Montréal challenges narratives of a French–Indigenous “admixed founding population”—narratives that continue to shape false-Indigeneity debates today. Our study of Notre-Dame Cemetery finds exclusively French European ancestry. doi.org/10.1073/pnas...
doi.org
Ancient DNA challenges to admixture narratives: Insights from Notre-Dame Cemetery (1683–1796) in Colonial Montreal, Canada | PNAS
Historical narratives of early colonization in Quebec frequently portray French–Indigenous relations as cooperative and characterized by admixture;...
052
Shai Carmi @shaicarmi.bsky.social · 03/10/2026
Thank you ❤️ My own vote will definitely support a new government!
010
Reposted by Shai Carmi
Pol Solé Navais @polsolenavais.bsky.social · 02/10/2026
📣📣📣 We are hiring a PhD student in reproductive genetics! The candidate will join our computational efforts to map the genetics behind pregnancy loss from both the fetal and maternal sides, working with a large pregnancy loss cohort and a new sequencing effort that we are launching. #PhD 1/3
web103.reachmee.com
Doctoral student in Medical Science - reproductive genetics and bioinformatics
Doctoral position in Medical Science Project title:  The role of aneuploidies and fetal and maternal genetics
135
Shai Carmi @shaicarmi.bsky.social · 02/10/2026
It is now October 2026. Since June 2025 (16 months), my kids have had only 7 months of normal school schedule. Since 2019 we have not had a single normal year. I am running on fumes.
120
Shai Carmi @shaicarmi.bsky.social · 24/09/2026
To be fair it's a terrible journal and it's been a mistake submitting there. But the current publication model makes zero sense. The current value of journals is negative. It's a burden on society.
011
Shai Carmi @shaicarmi.bsky.social · 23/09/2026
Just received from a journal an entirely AI-generated peer review. I knew the day would come. Academia is doomed.
152
Shai Carmi @shaicarmi.bsky.social · 22/09/2026
Given the recent decline of social media for high quality science discussions, I have recently moved most of my activity into this Slack group, dedicated to population genetics and genetic genealogy. Everyone is welcome to join: join.slack.com/t/geneticgen...
join.slack.com
Join Genetic genealogy science on Slack
Slack is a new way to communicate with your team. It’s faster, better organized, and more secure than email.
0167
Shai Carmi @shaicarmi.bsky.social · 21/09/2026
Interesting!
030
Shai Carmi @shaicarmi.bsky.social · 20/09/2026
Academic Age Acceleration: Your age minus your h-index (Mine's positive unfortunately)
031
Shai Carmi @shaicarmi.bsky.social · 20/09/2026
Indeed well powered. It seems solid, except that I didn't see any validation of the genotyping. Extremely rare variants are very difficult to genotype using an array (e.g. www.bmj.com/content/372/...). And unclear how/if it was imputed in chips not including it.
010
Shai Carmi @shaicarmi.bsky.social · 18/09/2026
Well, 23andMe and Harvard researchers just published a candidate SNP study in Science...
120
Reposted by Shai Carmi
alex ioannidis @alexgioannidis.bsky.social · 17/09/2026
Fst is a key genetic differentiation metric, but it emphasizes common variants over rare. In IEEE ICASSP out now lab member Margarita Geleta introduces Fq, which can be tuned across the allele freq spectrum. Fst & Shannon differentiation fall out of it as special cases. arxiv.org/abs/2511.03063
1124
Reposted by Shai Carmi
bioRxiv Genomics @biorxiv-genomic.bsky.social · 15/09/2026
Estimating de novo mutation rates using parent-offspring pairs www.biorxiv.org/content/10.64898/20…
062
Shai Carmi @shaicarmi.bsky.social · 13/09/2026
Paradoxically, the more fluent the writing, the more carefully I have to read to identify the nonsense. What a waste of our collective time.
020
Reposted by Shai Carmi
bioRxiv Genomics @biorxiv-genomic.bsky.social · 12/09/2026
Sparse Linear Algebra Accelerates Genotype Representation Graph Computation at Biobank Scale www.biorxiv.org/content/10.64898/20…
097
Shai Carmi @shaicarmi.bsky.social · 11/09/2026
One unrelated but amazing thing about this article is the reference to a "preprint published in August". Finally! Yes a preprint *is* a publication!
021
Reposted by Shai Carmi
Rémy Furrer @remyfurrer.bsky.social · 10/09/2026
What are IVF doctors’ views on polygenic embryo screening (PES/PGT-P), and what criteria shape their decisions? New survey results from 152 clinicians. @gimjournal.bsky.social @shaicarmi.bsky.social @toddlencz.bsky.social www.gimjournal.org/article/S109...
gimjournal.org
Decision-making criteria in polygenic embryo screening: A survey of reproductive medicine physicians
Polygenic embryo screening (PES) examines embryos for their genetic likelihood of developing complex conditions and traits. The commercialization of PES places reproductive endocrinologists and infert...
001
Reposted by Shai Carmi
Rémy Furrer @remyfurrer.bsky.social · 03/09/2026
Link to letter: www.economist.com/letters/2026... Link to study: jamanetwork.com/journals/jam...
economist.com
The opposition of scientific research on animals | Sep 5th 2026 Edition
Also this week, humanitarian aid, polygenic embryo selection, Dolly Parton, relaxing holidays, Lake Ontario
012
Reposted by Shai Carmi
alex ioannidis @alexgioannidis.bsky.social · 01/09/2026
Gnomix is published! www.nature.com/articles/s41... Still the most accurate & fastest local ancestry inference (LAI) method, bests Flare and others (RecombMix & Orchestra) released since our preprint. Easy to use, github.com/AI-sandbox/g... Use it with our toolkit, snputils.org
nature.com
Scalable high resolution ancestry deconvolution for genomic data - Nature Communications
The authors present Gnomix, a local ancestry framework that delivers leading accuracy across diverse admixed datasets on both whole-genome and array data with high efficiency, together with Gnofix, it...
22614
Reposted by Shai Carmi
Erkin Alaçamlı @masterolojist.bsky.social · 20/08/2026
Our new preprint is out: www.biorxiv.org/content/10.6... (1/n)
biorxiv.org
162
Reposted by Shai Carmi
Shai Carmi @shaicarmi.bsky.social · 12/08/2026
New preprint! Investigative genetic genealogy has revolutionized forensic identification, with hundreds of cases solved to date. But for any given new case, how likely is genetic genealogy to succeed? We used a whole-country genealogy to find out! 1/10 www.biorxiv.org/content/10.6...
biorxiv.org
Forensic investigative genetic genealogy match rate estimated from a nation-wide population register
Forensic investigative genetic genealogy (FIGG) is a revolutionary method in forensic genetics, whereby genetic relatives of an unknown target person are detected in direct-to-consumer genomic databas...
1124
Shai Carmi @shaicarmi.bsky.social · 12/08/2026
Great work by Harel Ismach, a student in my group, who has led the data analysis. Thanks to collaborators @gilig.bsky.social and @debbiekennett.bsky.social. The register data was made available by Israel's Central Bureau of Statistics.
031
Shai Carmi @shaicarmi.bsky.social · 12/08/2026
Nevertheless, our results are important because they apply to genealogical data *already accessible* to law enforcement in many countries, without relying on commercial providers. Our results should inform regulators and practitioners of FIGG, as well as quantify biobank privacy risks. 10/10
121
Shai Carmi @shaicarmi.bsky.social · 12/08/2026
Key limitations: * Finding a match does not guarantee identification * Many relatives missed because of the limited time depth of the register * Data errors (clerical, adoption, non-paternity, gamete donation, relationship misclassification, MZ twins) * Generalizability to other countries 9/10
121
Shai Carmi @shaicarmi.bsky.social · 12/08/2026
What is the probability of having *two* (or more) relatives in the database? This is important, as it is often the intersection of the family trees of multiple matches that leads to identification. The match rate naturally decreases, although not as much for the very large databases. 8/10
The probability of detecting multiple relatives of the target in the database. (A) The match rate when requiring at least two relatives in the database. All parameters are as in Figure 2. (B) The distribution of the number of relatives in the database. We only considered target individuals (alive or deceased) who have at least one relative in the database. The database size was 1% of the population (alive and >18) and we considered all 7th-degree or closer relatives. Target individuals already in the database count themselves towards the total number of relatives that they have there.
111
Shai Carmi @shaicarmi.bsky.social · 12/08/2026
Unexpectedly, the match rate barely increases beyond 5th-degree (2nd cousins). This is probably because the common ancestors of third cousins usually lived before the country's foundation. Implication: whenever relatives are found in the database, they are usually not too distant. 7/10
111
Shai Carmi @shaicarmi.bsky.social · 12/08/2026
Results: The match rate is the proportion of people with at least one relative (a "match") in the database. For a database covering 1% of the country, the match rate is ~25%. For a database covering 10% of the country, the rate is 68%. 6/10
The FIGG match rate using a national population register. We plot the proportion of the Israeli population (including deceased) who have at least one relative in the genomic database to whom they connect via a path in the population register of a maximal pre-defined length. The x-axis is the database size, in proportion of the living population over 18 years old. Each curve (legend) corresponds to a different maximal degree of relationship. For example, a first-degree relative is a parent or a full sibling, and a second-degree relative can be a half-sibling, a grandparent, or an aunt. Each data point is the average over 100 simulations. Standard deviations across simulations are shown as error bars, but are too small to be visible.
131
Shai Carmi @shaicarmi.bsky.social · 12/08/2026
FIGG simulation: We designated a random subset of the population as the "database". For each person in the database, we used the register to track all of their relatives up to 7th degree (eg, 3rd cousins). This gave us the list of people who would have one or more relatives in the database. 5/10
111
Shai Carmi @shaicarmi.bsky.social · 12/08/2026
Data: We used the entire Israeli national population register, documenting all parent-child relationships in all present and past citizens. After QC, the dataset covered 12.5 million people, among them 10.4 million alive. Summary stats look reasonable. 4/10
Characteristics of the IPIA data. (A) The distribution of birth decades (1870s-2010s). (B) The average number of children per individual (including childless individuals or individuals who have not yet completed their families) vs the parental birth decade. Birth decades start at the indicated year (e.g., the 1960 birth decade corresponds to birth years 1960-1969). (C) The distribution of the number of children per individual. We combined all individuals with 15 or more children into a single category (see also Figure S1). The y-axis is logarithmic. (D) The distribution of the paternal (light blue) and maternal age (pink) at childbirth, also known as the generation interval. The distribution is over all births in the database. Numbers on the x-axis indicate the midpoints of 3-year bins (e.g., 25 corresponds to ages 24-26).
131
Shai Carmi @shaicarmi.bsky.social · 12/08/2026
Our research question: Given a genomic database covering x% of the population of a country, what is the probability that a target person has one or more relatives in the database? Particularly, we focus on relatives from whom the target can be traced using available genealogical data. 3/10
111
Shai Carmi @shaicarmi.bsky.social · 12/08/2026
In forensic investigative genetic genealogy (FIGG), law enforcement people upload the DNA of an unknown person (a criminal or unidentified remains) to a consumer genomics database (eg FTDNA), find relatives, and intersect their family trees until reaching an identification. 2/10
A schematic of FIGG. For a subset of the population, a genome-wide DNA profile is available in a “genomic database” (light blue box). The identity of a target person (peach box) is unknown, but we have access to their genome-wide genetic profile. By comparing the DNA profile of the target against that of all individuals in the database, genetic relatives of the target (“matches”; red arrows) can be identified (light green box inside the database). Once matches have been detected, the investigation focuses on their relatives (green and red arrows), given that the target person must be one of them. Thus, to be identifiable using FIGG, a target person must have at least one relative in the database. We require the relationship to be equal or closer to a prespecified degree, corresponding to the most distant relationship that can be confidently detected using genetic data (usually 7th degree, e.g., a third cousin). In practice, more distant relationships can also sometimes be detected.
121
Shai Carmi @shaicarmi.bsky.social · 12/08/2026
New preprint! Investigative genetic genealogy has revolutionized forensic identification, with hundreds of cases solved to date. But for any given new case, how likely is genetic genealogy to succeed? We used a whole-country genealogy to find out! 1/10 www.biorxiv.org/content/10.6...
biorxiv.org
Forensic investigative genetic genealogy match rate estimated from a nation-wide population register
Forensic investigative genetic genealogy (FIGG) is a revolutionary method in forensic genetics, whereby genetic relatives of an unknown target person are detected in direct-to-consumer genomic databas...
1124
Shai Carmi @shaicarmi.bsky.social · 12/08/2026
Great work by Harel Ismach, a student in my group, who has led the data analysis. Thanks to collaborators @gilig.bsky.social and @debbiekennett.bsky.social. The register data was made available by Israel's Central Bureau of Statistics.
000
Shai Carmi @shaicarmi.bsky.social · 12/08/2026
Nevertheless, our results are important because they apply to genealogical data *already accessible* to law enforcement in many countries, without relying on commercial providers. Our results should inform regulators and practitioners of FIGG, as well as quantify biobank privacy risks. 10/10
110
Shai Carmi @shaicarmi.bsky.social · 12/08/2026
Key limitations: * Finding a match does not guarantee identification * Many relatives missed because of the limited time depth of the register * Data errors (clerical, adoption, non-paternity, gamete donation, relationship misclassification, MZ twins) * Generalizability to other countries 9/10
100
Shai Carmi @shaicarmi.bsky.social · 12/08/2026
What is the probability to have *two* (or more) relatives in the database? This is important, as it is often the intersection of the family trees of multiple matches that leads to identification. The match rate naturally decreases, although not as much for the very large databases. 8/10
The probability of detecting multiple relatives of the target in the database. (A) The match rate when requiring at least two relatives in the database. All parameters are as in Figure 2. (B) The distribution of the number of relatives in the database. We only considered target individuals (alive or deceased) who have at least one relative in the database. The database size was 1% of the population (alive and >18) and we considered all 7th-degree or closer relatives. Target individuals already in the database count themselves towards the total number of relatives that they have there.
100
Shai Carmi @shaicarmi.bsky.social · 12/08/2026
Unexpectedly, the match rate barely increases beyond 5th-degree (2nd cousins). This is probably because the common ancestors of third cousins usually lived before the country's foundation. Implication: whenever relatives are found in the database, they are usually not too distant. 7/10
100
Shai Carmi @shaicarmi.bsky.social · 12/08/2026
Results: The match rate is the proportion of people with at least one relative (a "match") in the database. For a database covering 1% of the population, the match rate is ~25%. For a database covering 10% of the pop, the rate is 68%. 6/10
The FIGG match rate using a national population register. We plot the proportion of the Israeli population (including deceased) who have at least one relative in the genomic database to whom they connect via a path in the population register of a maximal pre-defined length. The x-axis is the database size, in proportion of the living population over 18 years old. Each curve (legend) corresponds to a different maximal degree of relationship. For example, a first-degree relative is a parent or a full sibling, and a second-degree relative can be a half-sibling, a grandparent, or an aunt. Each data point is the average over 100 simulations. Standard deviations across simulations are shown as error bars, but are too small to be visible.
100
Shai Carmi @shaicarmi.bsky.social · 12/08/2026
FIGG simulation: We designated a random subset of the population as the "database". For each person in the database, we used the register to track all of their relatives up to 7th degree (eg, 3rd cousins). This gave us the list of people who would have one or more relatives in the database. 5/10
100
Shai Carmi @shaicarmi.bsky.social · 12/08/2026
Data: We used the entire Israeli national population register, documenting all parent-child relationships in all present and past citizens. After QC, the dataset covered 12.5 million people, among them 10.4 million alive. Summary stats look reasonable. 4/10
100
Shai Carmi @shaicarmi.bsky.social · 12/08/2026
Our research question: Given a genomic database covering x% of the population of a country, what is the probability that a target person has one or more relatives in the database? Particularly, we focus on relatives from whom the target can be traced using available genealogical data. 3/10
100
Shai Carmi @shaicarmi.bsky.social · 12/08/2026
In forensic investigative genetic genealogy (FIGG), law enforcement people upload the DNA of an unknown person (a criminal or unidentified remains) to a consumer genomics database (eg FTDNA), find relatives, and intersect their family trees until reaching an identification. 2/10
100
Shai Carmi @shaicarmi.bsky.social · 10/08/2026
It reminds me of when I TA'd physics to biology undergrads and the very first exercise I taught was to estimate the weight of all oceans.
030
Reposted by Shai Carmi
Mike Inouye @mikeinouye.bsky.social · 02/08/2026
If you're looking to do a postdoc or a PhD, check out our exemplar publications and feel free to get in touch: www.inouyelab.org/home/publica...
inouyelab.org
Inouye Lab - Publications
Google Scholar Inouye Lab authors * denotes equal contribution
013
Reposted by Shai Carmi
alex ioannidis @alexgioannidis.bsky.social · 30/07/2026
Just got our @genomeresearch.bsky.social article by Margarita Geleta. 📬 I love hard copies...still here's a link genome.cshlp.org/content/36/2... Rita's article was a tour de force of genomic uses for the autoencoder architecture: compression, dim reduction, ancestry, synthetic genomes.
1113
Reposted by Shai Carmi
Sini Nagpal @nagpalsini.bsky.social · 24/07/2026
Excited to share our work on "Pervasive interactions between exposures and polygenic risk can inform more effective clinical and behavioral interventions" @genomestake.bsky.social @natgenet.nature.com - www.nature.com/articles/s41...
131
Reposted by Shai Carmi
Nicola Pirastu @npirastu.bsky.social · 25/07/2026
We’re looking for a Statistical Geneticist to join our team at @humantechnopole.bsky.social. The role offers the opportunity to work across a wide range of projects, supporting research groups throughout Human Technopole while developing and implementing new statistical methods tinyurl.com/44uyt8u2
tinyurl.com
Statistical Geneticist
Statistical Geneticist
046
Reposted by Shai Carmi
loic-yengo.bsky.social @loic-yengo.bsky.social · 21/07/2026
Come and join our dream team! We’re recruiting a PhD candidate with a strong background in statistics, machine learning and (human) genetics to work at the interface of biological sequence foundational models and scores polygenic score methods.
2108
Reposted by Shai Carmi
Padraig Dixon @padraigdixon.bsky.social · 16/07/2026
New preprint: genetic variation and healthcare spending Across ~1.43M people, we identify hundreds of cost-associated variants, strong HLA signals, and higher costs among carriers of rare pathogenic variants in genes including BRCA1/2 and APC. www.medrxiv.org/content/10.6...
medrxiv.org
Quantifying the contribution of genetic variation to healthcare expenditure across diverse healthcare systems
Healthcare systems must balance rising costs with the delivery of effective care, yet the factors underlying large inter-individual differences in healthcare expenditure remain incompletely understood...
013