Pinned
May 7 •
Guidelines and Rules
This community exists for one reason — to make molecular biology genuinely accessible to people who need it for their work. Whether you're a data scientist working with genomic data, a researcher from an adjacent field, a science communicator, or someone transitioning into biology, you belong here. These guidelines exist to keep the community useful, respectful, and worth your time. Be curious, not performative There are no stupid questions here. If you don't understand something, ask. The whole point of this community is that biology can feel impenetrable from the outside — asking for clarity is exactly what this space is for. You will never be made to feel embarrassed for not knowing something. Be specific when you ask questions The more context you give, the better the answer you'll get. Instead of "I don't understand gene expression," try "I'm working with RNA-seq data and I'm not sure what normalisation method to use — can someone explain why this matters biologically?" Specific questions get specific, useful answers. Engage with the journal club Every week a real recent paper with clinical implications gets broken down here. Read it, ask questions, share what surprised you, push back if something doesn't make sense. The journal club is only as good as the conversation around it — your engagement makes it better for everyone. Self-promotion — one dedicated space only You're welcome to share your own work, papers, projects, or resources — but only in the weekly "Share Your Work" thread pinned at the top of the community. Unsolicited self-promotion posted anywhere else will be removed. This keeps the feed focused and useful. No misinformation Biology is a field where precision matters. If you share something, make sure it's accurate. If you're not sure, say so. If you see something that looks wrong, flag it respectfully rather than publicly calling it out — send a DM or tag me directly. Respect everyone's starting point Members here come from wildly different backgrounds — some have PhDs, some have never taken a biology class. Both are equally welcome. Do not condescend, do not gatekeep, and do not make anyone feel like their question is beneath the community. If you wouldn't say it in a professional meeting, don't say it here.
4
0
Pinned
May 7 •
Welcome to Biology Unlocked
Really glad you're here. This community exists because biology has a gap problem. Brilliant people from data science, engineering, chemistry, bioinformatics, and science communication are working with biological concepts every day — and nobody gave them a proper map. That's what Biology Unlocked is. Here's how to get started: Step 1 — Head to the Classroom Start with the Foundations course. It's designed for every background and every starting point. Lessons 1.1 and 1.2 are free previews — available to everyone before you dive into your specialist track. Step 2 — Choose your track Once you've completed Foundations, move into the track that fits your work: Biology for Data — if you work with biological datasets Biology for the Bench-Adjacent — if you work alongside biologists Biology for Communicators — if you write, edit, or report on biology Step 3 — Join the journal club Every week I break down a real recent research paper with clinical implications — methods, data, figures, and what it actually means. Live sessions are open to everyone. Recordings are available to Practitioner and Immersive members. Step 4 — Ask questions This is the most important step. Don't sit with a question you don't understand. Post it in the feed, bring it to a Q&A session, or DM me directly. There are no stupid questions here — only ones that haven't been answered yet. A note from me: I built this because I kept seeing the same gap after 10 years of editing 500+ manuscripts — people who were exceptional in their own fields, hitting a wall with biology. You're not behind. You just needed the right starting point. This is it. Welcome aboard. Akshi
6
0
13d •
Journal club #9 dropping tonight only on Skool!
For over a century, we have known cancer cells carry the wrong number of chromosomes. We never knew whether that chaos causes the cancer or just comes along for the ride. Al-Zahrani et al. finally answered this question in a new paper in Nature. The barrier was always scale. When a whole chromosome arm shifts, hundreds of genes change dose at once, and finding the few drivers among the passengers was near impossible. So the team built CRISPR-KOALA, a tool that switches genes off and on at the same time, inside a living mouse. That matters, because aneuploidy does both: it deletes some genes and duplicates others. They screened 3,752 genes and found 90 cancer drivers, with 81 being newly identified genes. Interestingly, 90% of them would have been invisible in a culture flask. They only acted as drivers inside a living tumour, with a real immune system and real oxygen gradients around them. And the century-old question finally has its answer: give the tumour those driver genes directly, and it no longer needs the chromosomal chaos. The chaos was never wreckage. It was how the tumour got what it needed. We read the whole paper in Journal Club tonight 7 PM! Session is free for everyone to watch so don’t miss this!
0
0
27d •
Biology Unlocked Article #10
I am a scientist. But as an editor, I enforce the one thing I hate most as a scientist. Formatting requirements. And don’t get me wrong. I love these formatting jobs. This is very calming work for me. In fact, every second manuscript I work on is a reformatting job: a paper rejected by one journal that now has to be reformatted for the guidelines of the next one. As a scientist, it is very easy to get stuck in that loop. Write, submit, desk rejection, reformat, submit, desk rejection, reformat, and on it goes until a journal finally accepts the paper. And the cost is real. Reformatting after desk rejections was estimated to waste about 230 million USD in 2021, with up to 2.5 billion projected through 2030 if nothing changes. Researchers spend around 1.55 million hours a year on reformatting alone, on top of writing manuscripts, chasing grants, running labs, and teaching. But I also sit on the other side. As someone from the publishing industry, I understand why journals want ready manuscripts. It streamlines review and shortens the time to a decision. Word count and section structure genuinely help you judge fit and rigor faster. But do references and layout do the same? They are cosmetic. They can be copy edited after acceptance. And a lot of journals already work this way. Elsevier’s Your Paper Your Way is a common example. So where is the line? This is the part I keep getting stuck on. I defend word count and structure as an editor, and I believe in them. But I cannot fully explain why those are sacred while reference style is disposable. Both are just packaging around the same science. So where would you draw the line? Are word count and structure a fair line that helps everyone, or just another habit we will eventually drop the way we are dropping reference styles? hashtag#AcademicPublishing hashtag#ScientificWriting hashtag#ResearchLife hashtag#PeerReview hashtag#ScholarlyPublishing
1
0
27d •
Journal Club #7 is happening today!
This is for all the bioinformaticians and data scientists in the group. Seven AI models were built to predict what a cell does when you switch off a gene. They were tested against a baseline that predicts nothing changes at all. And the baseline won. The models were not obscure: scGPT, scFoundation, scBERT, Geneformer, and UCE. Plus GEARS and CPA. Two of them claim perturbation prediction in their own papers. The baselines were deliberately embarrassing: one predicts zero change, one adds the two single effects together, and one returns the average across everything it has seen and ignores which gene you even asked about. None of the seven outperformed them. And the likely reason has nothing to do with machine learning. These models were pre-trained on observational data where millions of cells are just being cells. Nobody intervened. Then they were asked to predict an intervention: reach in, break this gene, and see what happens. That is a causal question. You cannot answer it from data where nobody ever intervened. The answer was never in the training set. This is where bench knowledge stops being optional. If you want to know what happens when you break something, you have to break it. Watching is not the same experiment. And the authors showed it. They pre-trained that same simple linear model on perturbation data instead of observational data, and it reliably beat everything else. This paper only exists because an anonymous peer reviewer, on an earlier paper from the same group, suggested in passing that they compare against a linear model. And the argument is not over. A preprint last October says the metrics were miscalibrated and the models do win once you score them properly. Notice what both sides agree on: you run the baseline. They are only fighting about the measuring stick. The title says these models do not YET outperform simple baselines. That word is doing work. This is not AI-bashing. It is a field being asked to check its floor before it celebrates its ceiling.
0
0
1-30 of 30
powered by
skool.com/biology-unlocked-7569
Molecular biology for data scientists, AI engineers, editors and research-adjacent professionals. No biology degree needed. Start from anywhere.
Suggested communities
Powered by