Semiha Kozan (@GeneToEvidence) header
Semiha Kozan (@GeneToEvidence) profile picture

Molecular Biology & Genetics | Bioinformatics & scientific writing | Genomics, health, AI—and the evidence behind the headlines.

Joined July 2026
Semiha Kozan retweeted
In vivo gene editing landscape. We mapped ~60 programs across 35 companies spanning DNA, RNA and epigenetic editors delivered by LNP, AAV and other vehicles. Some thoughts on this space: • In the last few years, gene editing has become a commercial reality (i.e. Casgevy), but that launch has shown how hard it is to scale an ex vivo approach: cell collection logistics, myeloablative conditioning, lag time to revenue recognition and more • Just as we're seeing in CAR-T (>$10B+ in acquisitions across Capstan, Orbital, Kelonia, Interius and EsoBiotec), there is hype around in vivo to remove a lot of those bottlenecks. Cells are edited inside the patient with a single off-the-shelf product, so there's far less manufacturing complexity • The field is early, with just two late-stage programs: Intellia's nex-z and Ionvo-z, which share the same LNP and Cas9 mRNA. The latter in particular will set the regulatory, payor and physician precedent for the field with a March 2027 PDUFA • The sell side is already modeling a launch ramp ~20% faster than ex vivo over the first 5 years, but a lot of this rests on physician comfort given the edit is permanent and chronic drugs already work • Since LNPs mostly reach the liver, there's commercial overlap between late-stage RNA drugs and these editing programs (HAE, ATTR-CM, lipids). For physicians, that creates an either / or scenario between a permanent edit and a proven chronic therapy, and many may just wait and see. • There are plenty of catalysts to watch through the end of 2026 and early 2027, including the Lonvo-z PDUFA, EDIT-401 human PoC and PM647 first data, among others And this is just a preview - tomorrow we’re dropping the full gene therapy landscape, covering every program across editing, AAV and non-viral delivery, plus deals, China and upcoming catalysts. Leave a note below and we'll send it to you when it drops.
18
43
6
214
17,166
Semiha Kozan retweeted
Graphic from @0xDeliriumm The Physics of Machine Reasoning
4
35
187
12,754
Last night an AI agent performed brain surgery on a mouse. We gave it 19 gene therapy vectors in a spreadsheet, a mouse model of epilepsy, access to a motorize surgery rig with cameras and one goal: reduce seizures. It picked the drug. It picked the brain region. It derived the coordinates from three papers. Then it drove the stereotaxic arms into the brain and injected 250 nanolitres while watching the entry site through the rig camera for bleeding. A licensed vet gated every irreversible step and held a physical kill switch to stop the AI agent. Thread on what it decided, what it saw, and the moment it refused to let me override it. [Writing this on the spot out of excitement, but I've been 18hrs in front of my laptop coding and coordinating this, so keeping it brief and descriptive, more cohesive thoughts on this soon...]
311
1,293
699
9,694
1,686,724
Semiha Kozan retweeted
Today, in collaboration with Anthropic, we’re excited to share rapid initial progress towards an antibody treatment for food allergy, which affects over 200 million people worldwide, discovered with the help of Claude. At Divergence Labs, we believe that AI can unlock the next generation of antibody therapeutics with more effective mechanisms than before. Such mechanisms are urgently needed in food allergy, which involve overcoming one of the tightest affinity protein pairs in immunology, IgE and its receptor FcεRI, on mast cells. Treatment with Xolair—the only approved antibody for food allergy—sequesters circulating IgE, but mast cells remain sensitized for weeks to months due to the extraordinarily slow dissociation of IgE from FcεRI. We envision a treatment that actively pulls IgE off of FcεRI, without waiting for natural dissociation. Like picking a lock, this requires the right combination of geometry and force. Our research team, working with Claude, ideated on mechanisms to accomplish this disruption, and converged on an antibody format with two arms with different functions and epitopes. Within one month, we designed, synthesized, and experimentally confirmed the biophysical function of the two arms, which we are now assembling into full therapeutic formats.
133
630
130
5,136
510,998
Glycosylation is among the most important modifications in biology. We know relatively little about glycobiology, because it is hard to study. This Nat Rev Methods primer describes our best tools for analyzing glycosylation 🔽
5
68
3
389
17,688
Semiha Kozan retweeted
Anthropic's modelling stack writes ribosome scripts that run the translation instead of animating it Most ribosome visuals are choreographed, which means the animation can only ever show what someone already believed. Here the ribosome is a script with rules, and what you watch is whatever those rules produce when you feed it a real sequence: > Encode: each step is written as a condition, not a keyframe > Feed: an actual mRNA sequence goes in, codon by codon > Run: the ribosome moves because the rules fire, not because a timeline says so • What emerges that nobody animated > It stalls on rare codons, exactly where the tRNA pool runs thin > A frameshift happens on its own when the sequence sets one up > Two ribosomes on one transcript collide, because nothing stopped them • Why a script beats an animation > Change the sequence and you get a different movie, not a re-render request > A behaviour that surprises you is a result, not an artistic choice > The same script runs 10,000 sequences overnight and reports the odd ones • Where it stays honest > Every rule traces to a measured rate, not a guess > A step with no rate in the literature refuses to run silently > The output logs which rules fired, so a weird result is debuggable One sequence translates in real-time-equivalent in about 40 seconds. And the part that broke my brain: Every animation I'd seen showed translation running at one smooth constant speed. Fed a real gene, the script stalled three times and one of the stalls is documented. Same process. One version could only show what its animator knew. The hard part was never drawing the ribosome. It was letting it do something nobody scripted ↓
12
59
2
468
26,144
Semiha Kozan retweeted
93 года спустя стало понятно , кто был прав — Бор или Шрёдингер? Осенью 1926 года Эрвин Шрёдингер приехал в Копенгаген защищать свою волновую механику. Нильс Бор встретил его жёстко. Бор утверждал: электрон прыгает между уровнями энергии мгновенно. Никакого пути. Никакого промежуточного состояния. Шрёдингер отвечал: нет. Он течёт непрерывно. «Если эта чёртова квантовая прыгучесть останется навсегда, — сказал Шрёдингер, — мне жаль, что я вообще связался с квантовой теорией». Они спорили несколько дней. Шрёдингер заболел и слёг с температурой прямо в доме Бора. Бор сел на край кровати и продолжил спор. Его ответ был коротким: «А мы все благодарны, что вы всё-таки связались». В видео ниже — не метафора. Это реальные решения уравнения Шрёдингера для атома водорода. Каждая форма — один электрон, находящийся одновременно в двух состояниях и плавно перетекающий из одного в другое. Никакого прыжка. Только непрерывная эволюция во времени. В 2019 году физики из Йеля поймали квантовый скачок «на полпути». Они показали: если наблюдать достаточно быстро и непрерывно, переход не мгновенный. Он занимает время. И его можно даже остановить и повернуть обратно. Но при обычном измерении электрон всегда оказывается целиком на одном уровне. Никогда «наполовину». Бор был прав насчёт того, что мы видим, когда смотрим. Шрёдингер был прав насчёт того, что происходит, пока мы не смотрим. Квантовый мир не выбирает одну картину. Он держит обе сразу — до момента измерения. Именно это и делает его таким странным и красивым. 👇Анимация — Imagining Physics
19
180
6
734
38,365
Semiha Kozan retweeted
Look at these gorgeous, dynamic, orange mitochondria surrounding the blue nucleus. A view of life from beneath a cell monolayer in a parti dish. Whatever angle you look at it, life is always dynamic and pretty special.
44
327
40
1,466
46,173
Semiha Kozan retweeted
What if immune cells could be reprogrammed to fight cancer directly inside the body, instead of in a lab? A new @NatureBiotech study from MIT used lipid nanoparticles to deliver CAR mRNA directly to natural killer (NK) cells. Up to 80% of NK cells were reprogrammed in mice using an NKG2D-targeted particle. Targeting NKG2A, an inhibitory checkpoint, both delivered the mRNA and relieved an inhibitory brake, leading to complete tumor regression in most treated mice in a melanoma model, with no recurrence through day 100. CAR expression was temporary, and repeat dosing may be needed. Still preclinical. But a cell therapy delivered from a vial instead of manufactured in a lab could transform access to cancer treatment.
8
29
3
197
26,735
1/ 2026 Nobel Tıp Ödülü ışıkla açılıp kapanan bir anahtara verildi: Karl Deisseroth, Peter Hegemann ve Georg Nagel, ışıkla açılan iyon kanalları ve optogenetik için (Nobel Assembly, 2026). #yapayzeka #beyin
8
77
7
441
26,328
Semiha Kozan retweeted
Aucun laboratoire au monde ne pourra jamais tester 9 milliards de mutations de l'ADN. Une IA de Google vient de toutes les prédire. AlphaGenome, c'est un réseau de neurones de DeepMind. Il avale jusqu'à un million de lettres d'ADN d'affilée, et il vous dit ce que change une mutation à cet endroit. Le tout en une seconde. DeepMind l'a lancé sur les 9 milliards de mutations possibles du génome humain, et vient de tout publier. Gratuit pour la recherche académique (oui, gratuit). À une seconde par mutation, ça fait 285 ans de calcul. Ils l'ont fait une fois, pour tout le monde. Je m'explique sur pourquoi c'est énorme. Séquencer un patient ne coûte presque plus rien et prend quelques jours. Le problème, c'est qu'on ressort avec des milliers de mutations inconnues et qu'il faut deviner laquelle rend le malade ... malade. Résultat : d'après une étude européenne sur 6 507 patients, il faut en moyenne 4,7 ans pour poser un diagnostic de maladie rare. Presque cinq ans d'errance, pour 300 millions de personnes dans le monde. Et ça marche ... une équipe du Broad Institute a identifié grâce à l'Atlas la cause probable d'une épilepsie sévère qu'on n'expliquait pas. Une mutation planquée dans les 98 % du génome qu'on ne savait pas lire. Les chiffres : → 9 milliards de mutations prédites, environ 27 000 prédictions chacune → 1 pétaoctet de données, plus de 30 fois AlphaFold → Score AVI : le bon coupable dans le top 50 dans 29,5 % des cas Ce dernier chiffre est celui qui compte. L'outil standard aujourd'hui, CADD, y arrive dans 12,5 % des cas. On passe d'une chance sur huit à une chance sur trois de mettre la bonne mutation sous les yeux du médecin. C'est la même équipe qui avait sorti AlphaFold et fait passer la biologie de 190 000 structures connues à plus de 200 millions. Ils refont exactement le même geste sur la génétique. Excellent non ? bravo Google.
We’re launching AlphaGenome Atlas: an AI-powered searchable database mapping the predicted impact of all 9 billion possible single-letter DNA changes. Here’s how it could help researchers better understand our biology 🧵
16
389
18
1,551
100,895
A single workout reduces overthinking. You don’t have to think your way out of the loop when you can move out of it.
65
918
88
7,030
576,741
Semiha Kozan retweeted
I want to know God's thoughts - the rest are mere details. — Albert Einstein
43
260
20
1,425
71,232
Semiha Kozan retweeted
metacognition, the ability to observe and understand your own thinking, is what neuroscience considers the highest form of intelligence. it is what separates reactive people from self aware ones. when you can step back and notice why you are thinking what you are thinking, you gain the ability to change it. most people live inside their thoughts. the rare few learn to watch them.
72
638
62
3,169
242,238
Semiha Kozan retweeted
Built an insane model of the human cell in just 30 minutes 🤯 GPT-6 Astra is ridiculously powerful. Can’t wait to put it to the test on biological tasks. @RimShayakhmetov @DeryaTR_ take a look!
89
221
42
1,668
193,718
Semiha Kozan retweeted
your brain is physically changing as you watch this right now. every thought, experience, and a behavior can influence which neural connections are strengthened, weakened, or reorganized. this is neuroplasticity.
look closer. what you call "change" looks like this inside your brain:
12
72
2
424
20,678
Semiha Kozan retweeted
Bugün "yapay zekâ" teriminin ilk kez kullanılmasının 71. yıldönümü.
Over 70 years ago today, the term “artificial intelligence” was coined in a conference proposal: stanford.io/2WJJJGN
3
19
250
15,191
🚨 BREAKTHROUGH: A molecule in broccoli & cabbage just killed 90% of cavity-causing bacteria in lab studies. And no, it’s not fluoride. Here’s how this changes everything about dental care.🧵
21
470
18
3,521
265,293
Semiha Kozan retweeted
A single neuron does three things and nothing more: weigh its inputs, bend the result with a curve, then quietly shrink its own mistakes. z = Σᵢ wᵢxᵢ + b a = 1/(1 + e⁻ᶻ) ŷ = a That gentle S-shape is the entire reason deep networks can separate data that no straight line ever could. The same unit now flags tumors, pilots drones, and predicts markets. It has one goal: make the average distance between ŷ and truth as small as possible. Linear algebra, then a nonlinearity. That is the art.
58
576
41
2,864
110,632