Lightnews — Scholar-powered news

Reposted by Babak Alipanahi

Andrea Love, PhD | Biomedical Scientist @drandrealove.bsky.social · 14d

The strongest predictor of autism is genetics.

Data:
• If one identical twin has ASD, the other has a 90–99% chance.
• Compare that to 31–67% chance for fraternal twins.
• Non-twin siblings have ~20% likelihood of ASD, 7-fold higher than general population.

1/

9 71 260

Babak Alipanahi @babak-a.bsky.social · 14d

Wow...what? Apple does protein folding now? As the name implies, it is a "simple" and general-purpose transformer-based method, which relies on a non-domain-specific flow-matching objective rather than domain knowledge.

AI x Bio Discovery @aixbiobot.bsky.social · 15d

SimpleFold: Folding Proteins is Simpler than You Think [new]
Folding Proteins is Simpler than You Think
...by using a flow-matching objective and general-purpose transformers, achieving competitive results without complex domain-specific components.

1

Babak Alipanahi @babak-a.bsky.social · 16d

Such an amazing method and such a good study! Have to read it more carefully, but shows huge potential. Combining evolutionary and population genomics + language model is a great idea.

Yun S. Song @yun-s-song.bsky.social · 17d

We are excited to share GPN-Star, a cost-effective, biologically grounded genomic language modeling framework that achieves state-of-the-art performance across a wide range of variant effect prediction tasks relevant to human genetics.
www.biorxiv.org/content/10.1...
(1/n)

3

Reposted by Babak Alipanahi

Jonathan Wosen @jonathanwosen.bsky.social · Aug 20

I'm a science journalist for @statnews.com looking to talk with folks who applied to biomedical Ph.D. programs and had offers rescinded earlier this year. What have you been up to since then? You can reach me at [email protected] or via Signal (username jwosen.27). #journorequest

55 55

Babak Alipanahi @babak-a.bsky.social · Aug 19

Nothing beats the elegance and simplicity of AUC with confidence intervals. AUPRC is useful, but it looks jagged and the baseline is cohort-specific.

1

Babak Alipanahi @babak-a.bsky.social · Aug 15

I am not sure why a blog post is paywalled.

1 1

Reposted by Babak Alipanahi

Constantin Ahlmann-Eltze @const-ae.bsky.social · Aug 4

Our paper benchmarking foundation models for perturbation effect prediction is finally published 🎉🥳🎉

www.nature.com/articles/s41...

We show that none of the available* models outperform simple linear baselines. Since the original preprint, we added more methods, metrics, and prettier figures!

🧵

Beeswarm plot of the prediction error across different methods of double perturbations showing that all methods (scGPT, scFoundation, UCE, scBERT, Geneformer, GEARS, and CPA) perform worse than the additive baseline.

Line plot of the true positive rate against the false discovery proportion showing that none of the methods is better at finding non additive interactions than simply predicting no change.

2 57 130

Babak Alipanahi @babak-a.bsky.social · Aug 4

This is a pretty stark reminder that every methods paper needs proper baselines: "We compared five foundation and two deep learning models against deliberately simple baselines for predicting transcriptome changes after single or double perturbations. None outperformed the baselines."

Nature Methods @natmethods.nature.com · Aug 4

An analysis shows that current deep learning models do not beat linear baselines in predicting gene perturbation effects, thus emphasizing the importance of further method development and evaluation. @const-ae.bsky.social @wkhuber.bsky.social @s-anders.bsky.social

www.nature.com/articles/s41...

Deep-learning-based gene perturbation effect prediction does not yet outperform simple linear baselines - Nature Methods

The analysis presented in this Brief Communication shows that, despite their complexity, current deep learning models do not outperform linear baselines in predicting gene perturbation effects, thus e...

www.nature.com

1

Babak Alipanahi @babak-a.bsky.social · Aug 1

RNA for the win! The once-overlooked black sheep of the macromolecules is slowly yet surely claiming its rightful place and prominence.

Institute of Molecular Genetics @imgprague.bsky.social · Aug 1

🧬 Happy RNA Day! 🎉
A tiny molecule with a massive impact #RNA!
From translating genetic code to fighting diseases through mRNA vaccines, RNA is currently at the heart of innovation in science and medicine.

3

Reposted by Babak Alipanahi

Institute of Molecular Genetics @imgprague.bsky.social · Aug 1

🧬 Happy RNA Day! 🎉
A tiny molecule with a massive impact #RNA!
From translating genetic code to fighting diseases through mRNA vaccines, RNA is currently at the heart of innovation in science and medicine.

3 21

Babak Alipanahi @babak-a.bsky.social · Jul 29

Didn't know Ben Blencowe has a part-time appointment at Crick! Should be a productive collaboration!

The Francis Crick Institute @crick.ac.uk · Jul 29

Meet Jernej Ule (@ulelab.bsky.social) and Ben Blencowe, who set up the Brain RNA Regulatory Networks satellite lab here at the Crick.

We spoke to Ben and Jernej about their research into RNA auto-gated vectors and their potential applications in gene therapy.

www.crick.ac.uk/news/2025-07...

Introducing... Jernej Ule and Ben Blencowe

We spoke to Jernej Ule and Ben Blencowe, who recently set up the Brain RNA Regulatory Networks satellite lab here at the Crick.

www.crick.ac.uk

Reposted by Babak Alipanahi

Nature Biotechnology @natbiotech.nature.com · Jul 25

Predicting the translation efficiency of messenger RNA in mammalian cells go.nature.com/4mgXR0C

Predicting the translation efficiency of messenger RNA in mammalian cells - Nature Biotechnology

A deep convolutional neural network model predicts the influence of the full-length mRNA sequence on translation efficiency.

go.nature.com

1 12 42

Babak Alipanahi @babak-a.bsky.social · Jul 17

I asked ChatGPT to summarize this article and this is what it came up with: "Overall: AI itself isn’t the problem—it’s the hype, incentives, lack of accountability, and cultural shifts that could slow meaningful science."

I think it is a pretty good summary.

2 1

Babak Alipanahi @babak-a.bsky.social · Jul 16

@atulbutte.bsky.social's last paper. He will be dearly missed.

AI x Bio Discovery @aixbiobot.bsky.social · Jul 16

Auto-MedCalc: Automated Biomarkers Discovery and Risk Score Generation with AI Agents [new]
Uses AI agents to automate biomarker discovery and risk score generation from multi-source biomedical data for improved diagnosis.

1

Reposted by Babak Alipanahi

Health Nerd @gidmk.bsky.social · Jul 7

How did the pandemic start?

I have avoided publicly commenting on this for 5 years. Here's my take as someone who has no stake at all in the game.

gidmk.substack.com/p/pandemic-r...

Pandemic Revisionism, Part One: COVID-19 Origins

Because people keep getting this very wrong.

gidmk.substack.com

20 120 350

Reposted by Babak Alipanahi

Eric Topol @erictopol.bsky.social · Jul 6

Summary Table of the massive defunding of biomedical research in the United States
www.nytimes.com/interactive/...

60 1K 1.6K

Babak Alipanahi @babak-a.bsky.social · Jun 27

Amgen’s MariTide (GLP‑1 agonist + GIP antagonist), a once‑monthly injection, is demonstrating strong efficacy. The mean percentage change in body weight from baseline to week 52 ranged from −12.3% to −16.2%. Impressive results!

www.nejm.org/doi/full/10....

Once-Monthly Maridebart Cafraglutide for the Treatment of Obesity — A Phase 2 Trial | NEJM

Maridebart cafraglutide (known as MariTide) is a long-acting peptide–antibody conjugate that combines glucagon-like peptide-1 receptor agonism and glucose-dependent insulinotropic polypeptide recep...

www.nejm.org

Reposted by Babak Alipanahi

Wellcome Sanger Institute @sangerinstitute.bsky.social · Jun 26

‘Huge advances in cancer and rare diseases’: 25 years of the human genome 🧬

Our Director, Matt Hurles, spoke to the Guardian @scienceweekly.bsky.social podcast about the Human Genome Project, and what could be in store over the next 25 years 🎧

www.theguardian.com/science/audi...

‘Huge advances in cancer and rare diseases’: 25 years of the human genome – podcast

It’s been 25 years since Bill Clinton announced one of humanity’s most important scientific achievements: the first draft of the human genome. At the time, there was a great deal of excitement about t...

www.theguardian.com

5 28

Reposted by Babak Alipanahi

Brian Naughton @btnaughton.bsky.social · Jun 26

Great article on cancer vaccines by @mackenziejem.bsky.social
www.mackenziemorehead.com/the-failures...

The Failures and Futures of Cancer Vaccines

The cancer vaccine field has yet to become a systematic science despite decades of work and a steadily rising number of clinical trials since the first clinical PoC a decade ago. A core problem faci...

www.mackenziemorehead.com

1 2

Reposted by Babak Alipanahi

EMBL-EBI @ebi.embl.org · Jun 26

The first draft of the human genome turns 25 today! 🎂

On 26 June 2000, the UK & US announced the first full map of the 3 billion DNA base pairs that make us human.

🌍 A milestone for humanity, powered by global collaboration.
www.youtube.com/watch?v=pnWA...

#HumanGenome #ScienceHistory 🖥️🧬

The Human Genome Project at 20 - interview with Ewan Birney

YouTube video by European Bioinformatics Institute - EMBL-EBI

www.youtube.com

1 15 47

Reposted by Babak Alipanahi

Elizabeth Warren @warren.senate.gov · Jun 17

Allowing Netanyahu to drag us into another endless war in the Middle East would be a catastrophic error by President Trump and Republicans in Congress.

Every lawmaker needs to ring the alarm against U.S. military action in Iran. Only Congress has the authority to declare war.

120 580 2.6K

Reposted by Babak Alipanahi

Lior Pachter @lpachter.bsky.social · Jun 16

I wrote a review of a recent paper on false discovery and multiple testing correction. liorpachter.wordpress.com/2025/06/16/r...

Reply to: Reply to: False positives in the study of memory-related gene expression

In the Nature paper “Spatial transcriptomics reveal neuron–astrocyte synergy in long-term memory” published on March 14th, 2024, authors Sun et al. claimed to identify cell-type specifi…

liorpachter.wordpress.com

7 40 150

Reposted by Babak Alipanahi

Our World in Data @ourworldindata.org · Jun 6

Less than 200 years ago, one in three Dutch children died before the age of five

A line graph illustrating the estimated share of newborns in the Netherlands who die before reaching age five from 1840 to 2023. The y-axis represents the percentage of children dying before age five, ranging from 0% to 40%, while the x-axis shows the years from 1840 to 2023.

The graph has a prominent downward trend, indicating a significant decline in child mortality rates over time. Key data points are highlighted: in 1840, around 40% of children died before age five, peaking around 1959, where 39% of Dutch children died before this age. A notable decrease begins after 1960, reaching about 10% in 1945, where 1 in 10 children died. In the latest data for 2023, this figure has improved to 1 in 250 children dying before reaching age five.

Annotations on the graph provide contextual information, linking past statistics to present outcomes. The data sources listed at the bottom include Gapminder (2015) and UN IGME (2025). The design is attributed with a copyright notice indicating CC BY

5 21 98

Babak Alipanahi @babak-a.bsky.social · Jun 5

These are the phenotypes they claim to screen:

list of phenotypes that Nucleus Genomics claims to be screening, including appearance, physical health, women's health and so on.

2

Babak Alipanahi @babak-a.bsky.social · Jun 4

I developed accurate polygenic risk scores at 23andMe using its huge database and can attest to their limitations, I see this craze stemming from a deep ignorance about this topic.
@lpachter.bsky.social wrote an amazing piece on this back in 2021: liorpachter.wordpress.com/2021/04/12/t...

The amoral nonsense of Orchid’s embryo selection

If you haven’t heard about Clubhouse yet… well, it’s the latest Silicon Valley unicorn, and the popular new chat hole for thought leaders. I heard about it for the first time a fe…

liorpachter.wordpress.com

1 3 13