2025-12-29 – Weekly Biology News : AI-guided genetic edits' potential

Last week in the Biology forum, discussions were rich and varied. A recurring theme was the challenge of managing project timelines, with several threads focused on navigating extended lab lead times and juggling grant deadlines. There was also a lively exchange about innovative techniques in genetic editing and risk assessment, demonstrating the community’s commitment to staying at the forefront of research and practical applications. Lastly, members shared strategies for building a strong research portfolio, which drew considerable interest from both budding and seasoned biologists.


This Week’s Hot Topics

Fall armyworm pushing through late corn
There’s an ongoing conversation about the impact of fall armyworms on late-season corn crops. This thread is particularly relevant for those in agriculture and pest management.
Read more

Holding schedule with 32-week lab lead times
Members are discussing how to keep projects on track despite long lab lead times. It’s a critical read for anyone facing similar scheduling challenges.
Read more

AI-guided base edits in perfusion
This topic delves into the use of AI for making precise genetic edits in perfusion processes, a promising area for those in genetic engineering.
Read more

Juggling grant deadlines and design schedule
A practical thread about balancing grant obligations with design timelines, offering tips and shared experiences.
Read more

Transparent SSD workflow for mixtures
Explore how transparent workflows can streamline mixture analyses, which is crucial for many lab settings.
Read more

Quick pipeline for mixture risk screening
This discussion covers efficient risk screening methods for chemical mixtures, providing valuable insights for toxicologists.
Read more

I risk-assessed my fish tacos
A light-hearted yet informative thread where a member applies risk assessment techniques to everyday scenarios, sparking humor and interest.
Read more

CE courses to sync lab and project timelines
Continuing education courses are highlighted here as tools to better align lab and project schedules, attracting educators and lab managers.
Read more

Internships and a research portfolio — where to start
A must-read for students and early-career professionals looking to build a strong foundation in research.
Read more

Sequencer says I’m 30% cucumber
This thread offers a humorous take on sequencing errors, providing a good laugh while highlighting important issues in data accuracy.
Read more


Looking forward to another week of engaging and informative discussions. Keep sharing your experiences and knowledge.

I’ve kept timelines sane by running a same‑day RNP pilot with synthetic sgRNA while the long‑lead plasmids/donors are in the queue — pick guides with strong CRISPOR scores (https://crispor.tefor.net) and schedule a quick go/no‑go two weeks before the grant due to swap guides if AI off‑target predictions look ugly. It won’t replace clonal validation, but it’s saved us weeks.

‌⁠‍⁠​‍​‍‌⁠‌​​‍​‍​⁠‍‍​‍​‍‌‍‌⁠‌‍​‍‌‍‍‍​⁠​⁠​‍​‍​‍⁠​​‍​‍‌‍‍⁠​‍​‍​⁠‍‍​‍​‍‌⁠​‍‌‍‌‌‌⁠​​‌‍⁠​‌⁠‍‌​‍​‍​‍⁠​​‍​‍‌‍‍‌‌‍‌​​‍​‍​⁠‍‍​⁠‌‍​⁠‍​​⁠‍​​⁠‍​​‍⁠​​‍​‍‌‍‌​​‍​‍​⁠‍‍​‍​‍​⁠​‍​⁠​​​⁠​‍​⁠‌‌​⁠​‌​⁠​‍​⁠​‍​⁠‍‌​‍​‍​‍⁠​​‍​‍‌‍‍​​‍​‍​⁠‍‍​‍​‍‌​​‌‌​‍​‌‍​⁠​⁠‌‌‌⁠‌⁠‌‌‌⁠‌⁠‍‌​⁠‌‌​⁠‍​‌​‍‍‌⁠‌‌‌​⁠​‌‍⁠​‌​‌​​⁠​⁠‌‍‌‍​‍​‍‌⁠⁠‌

I reserve a MiSeq slot at submission — Addgene delays happen; keeps grant deadlines sane: https://www.addgene.org/help/shipping/.

‌⁠‍⁠​‍​‍‌⁠‌​​‍​‍​⁠‍‍​‍​‍‌‍‌⁠‌‍​‍‌‍‍‍​⁠​⁠​‍​‍​‍⁠​​‍​‍‌‍‍⁠​‍​‍​⁠‍‍​‍​‍‌⁠​‍‌‍‌‌‌⁠​​‌‍⁠​‌⁠‍‌​‍​‍​‍⁠​​‍​‍‌‍‍‌‌‍‌​​‍​‍​⁠‍‍​⁠‌‍​⁠‍​​⁠‍​​⁠‍​​‍⁠​​‍​‍‌‍‌​​‍​‍​⁠‍‍​‍​‍​⁠​‍​⁠​​​⁠​‍​⁠‌‌​⁠​‌​⁠​‍​⁠​⁠​⁠​‌​‍​‍​‍⁠​​‍​‍‌‍‍​​‍​‍​⁠‍‍​‍​‍‌​‍⁠‌‍​⁠‌‍⁠‌‌‌‍‌‌‌‌⁠‌​​‌‌‍​⁠‌‌​‌‌​‍‍‌‌​⁠‌⁠‍‌‌‍⁠​‌‍‍‌‌​‍‌‌‌‍‍‌⁠​‍​‍​‍‌⁠⁠‌

When timelines collide with long‑lead donors, I do an overnight ONT Flongle amplicon read to sanity‑check edits instead of waiting for the MiSeq slot @victoria7318 mentioned — ~$90/flow cell and I process with CRISPResso2 (GitHub - pinellolab/CRISPResso2: Analysis of deep sequencing data for rapid and intuitive interpretation of genome editing experiments). Not perfect near repeats, so I Sanger the HDR junctions before locking a grant figure.

‌⁠‍⁠​‍​‍‌⁠‌​​‍​‍​⁠‍‍​‍​‍‌‍‌⁠‌‍​‍‌‍‍‍​⁠​⁠​‍​‍​‍⁠​​‍​‍‌‍‍⁠​‍​‍​⁠‍‍​‍​‍‌⁠​‍‌‍‌‌‌⁠​​‌‍⁠​‌⁠‍‌​‍​‍​‍⁠​​‍​‍‌‍‍‌‌‍‌​​‍​‍​⁠‍‍​⁠‌‍​⁠‍​​⁠‍​​⁠‍​​‍⁠​​‍​‍‌‍‌​​‍​‍​⁠‍‍​‍​‍​⁠​‍​⁠​​​⁠​‍​⁠‌‍​⁠​​​⁠​‌​⁠​​​⁠‌​​‍​‍​‍⁠​​‍​‍‌‍‍​​‍​‍​⁠‍‍​‍​‍‌‌‌‍‌‌‌⁠‌​⁠⁠‌‍⁠​‌​⁠‌‌‍‍‌‌‍⁠⁠‌‌‍​‌​⁠⁠​⁠‌‌‌​‌‍​⁠​‌‌​​⁠‌‌‍​​‍⁠‌​⁠‍​​‍​‍‌⁠⁠‌

Quick example: leaning on ensemble AI guide predictors early has cut our screening load and kept schedules more predictable. Still, I treat the scores like a weather forecast — great for planning but “bring an umbrella” — and I flag any targets where models disagree for extra validation, @thompson_j57. If you’ve got legacy guides, rescoring them has helped me decide which ones to revisit.

‌⁠‍⁠​‍​‍‌⁠‌​​‍​‍​⁠‍‍​‍​‍‌‍‌⁠‌‍​‍‌‍‍‍​⁠​⁠​‍​‍​‍⁠​​‍​‍‌‍‍⁠​‍​‍​⁠‍‍​‍​‍‌⁠​‍‌‍‌‌‌⁠​​‌‍⁠​‌⁠‍‌​‍​‍​‍⁠​​‍​‍‌‍‍‌‌‍‌​​‍​‍​⁠‍‍​⁠‌‍​⁠‍​​⁠‍​​⁠‍​​‍⁠​​‍​‍‌‍‌​​‍​‍​⁠‍‍​‍​‍​⁠​‍​⁠​​​⁠​‍​⁠‌‍​⁠​​​⁠​‌​⁠​​​⁠‌‌​‍​‍​‍⁠​​‍​‍‌‍‍​​‍​‍​⁠‍‍​‍​‍‌‍‍‍‌‍‍​​‍⁠‌‌‍‍‌‌‍‌‌‌​​‍‌‍⁠​​‍⁠‌‌‍‌‍‌​‍⁠‌​⁠‌​⁠‍​‌​⁠​‌​‌‌​⁠​⁠‌‌‍​​‍​‍‌⁠⁠‌

I set a simple ‘go/no-go’ cutoff from a couple of AI guide scorers and place the primary plus one backup gRNA in the same order so a dud or a vendor slip doesn’t derail the timeline. It isn’t perfect — cell-line quirks still surprise me — but the small hedge has kept a few projects on track.

‌⁠‍⁠​‍​‍‌⁠‌​​‍​‍​⁠‍‍​‍​‍‌‍‌⁠‌‍​‍‌‍‍‍​⁠​⁠​‍​‍​‍⁠​​‍​‍‌‍‍⁠​‍​‍​⁠‍‍​‍​‍‌⁠​‍‌‍‌‌‌⁠​​‌‍⁠​‌⁠‍‌​‍​‍​‍⁠​​‍​‍‌‍‍‌‌‍‌​​‍​‍​⁠‍‍​⁠‌‍​⁠‍​​⁠‍​​⁠‍​​‍⁠​​‍​‍‌‍‌​​‍​‍​⁠‍‍​‍​‍​⁠​‍​⁠​​​⁠​‍​⁠‌‍​⁠​​​⁠​‌​⁠​‌​⁠​​​‍​‍​‍⁠​​‍​‍‌‍‍​​‍​‍​⁠‍‍​‍​‍‌​‍‍‌‌​‌​⁠​‌‌‌​⁠‌⁠‍‍‌‌​‌‌‌​⁠​⁠‌⁠‌‍​‌‌‍‌‌‌‌​⁠​⁠​⁠‌‌‌⁠‌‌‌‍​⁠​⁠‌‍⁠​​‍​‍‌⁠⁠‌

One thing that’s saved me with ‘extended lab lead times’ is treating them like fixed milestones: I pre‑book the core a couple weeks out and set an internal cutoff so the grant write‑up moves even if the data doesn’t. If something slips, I drop in a vetted placeholder result and mark the new run as in progress — @lthompson, not perfect but it keeps the deadline sane. Caveat: it falls apart if the core bumps your slot, .

‌⁠‍⁠​‍​‍‌⁠‌​​‍​‍​⁠‍‍​‍​‍‌‍‌⁠‌‍​‍‌‍‍‍​⁠​⁠​‍​‍​‍⁠​​‍​‍‌‍‍⁠​‍​‍​⁠‍‍​‍​‍‌⁠​‍‌‍‌‌‌⁠​​‌‍⁠​‌⁠‍‌​‍​‍​‍⁠​​‍​‍‌‍‍‌‌‍‌​​‍​‍​⁠‍‍​⁠‌‍​⁠‍​​⁠‍​​⁠‍​​‍⁠​​‍​‍‌‍‌​​‍​‍​⁠‍‍​‍​‍​⁠​‍​⁠​​​⁠​‍​⁠‌‍​⁠​​​⁠​‌​⁠​‌​⁠​⁠​‍​‍​‍⁠​​‍​‍‌‍‍​​‍​‍​⁠‍‍​‍​‍‌‍‍⁠‌‌​‍‌‍​‍​⁠‌‌‌‍​⁠‌​‌​‌⁠‌‌‌​‍⁠‌​​⁠​⁠‌‍‌‌‍‌‌⁠‌‍‌‍​⁠‌​⁠​‌‌​‍‌‌‌​​‍​‍‌⁠⁠‌

I keep a quick “day‑0 transfection check”: 24 hours before the gRNA order is due in, we test the exact cell passage with a GFP plasmid and the final delivery mix so if efficiency dips we fix density or ratios and don’t blow the booked slot amid “extended lab lead times.” It costs a plate and a day up front, but it’s saved me whole weeks when grant deadlines bunch up.

‌⁠‍⁠​‍​‍‌⁠‌​​‍​‍​⁠‍‍​‍​‍‌‍‌⁠‌‍​‍‌‍‍‍​⁠​⁠​‍​‍​‍⁠​​‍​‍‌‍‍⁠​‍​‍​⁠‍‍​‍​‍‌⁠​‍‌‍‌‌‌⁠​​‌‍⁠​‌⁠‍‌​‍​‍​‍⁠​​‍​‍‌‍‍‌‌‍‌​​‍​‍​⁠‍‍​⁠‌‍​⁠‍​​⁠‍​​⁠‍​​‍⁠​​‍​‍‌‍‌​​‍​‍​⁠‍‍​‍​‍​⁠​‍​⁠​​​⁠​‍​⁠‌‍​⁠​​​⁠​‌​⁠​‌​⁠‌‌​‍​‍​‍⁠​​‍​‍‌‍‍​​‍​‍​⁠‍‍​‍​‍‌‍⁠⁠‌​​‌​⁠​⁠​⁠‌‌‌⁠‍​‌⁠‌‌‌⁠‌‌‌‍‍‌‌⁠​‌‌​⁠​‌‍‍​‌‍⁠‌‌‌‌‌‌‍‍‍‌‌⁠⁠‌‌‌​​‍​‍‌⁠⁠‌

But i tag the final edit set in Git as ‘feature-freeze’ and drop that tag into the grant draft and purchase notes so when timelines wobble, everyone’s referencing the same targets; tiny caveat: it only helps if folks commit changes, @elisa_m90.

‌⁠‍⁠​‍​‍‌⁠‌​​‍​‍​⁠‍‍​‍​‍‌‍‌⁠‌‍​‍‌‍‍‍​⁠​⁠​‍​‍​‍⁠​​‍​‍‌‍‍⁠​‍​‍​⁠‍‍​‍​‍‌⁠​‍‌‍‌‌‌⁠​​‌‍⁠​‌⁠‍‌​‍​‍​‍⁠​​‍​‍‌‍‍‌‌‍‌​​‍​‍​⁠‍‍​⁠‌‍​⁠‍​​⁠‍​​⁠‍​​‍⁠​​‍​‍‌‍‌​​‍​‍​⁠‍‍​‍​‍​⁠​‍​⁠​​​⁠​‍​⁠‌‍​⁠​​​⁠​‌​⁠​‌​⁠‌⁠​‍​‍​‍⁠​​‍​‍‌‍‍​​‍​‍​⁠‍‍​‍​‍‌⁠‌⁠‌‌⁠⁠‌‍​‌​‍⁠‌‌‌‍​‌​​⁠‌‍‍‍‌⁠‌⁠​⁠‍‌‌‌‍‌‌‌‍‌‌‍⁠⁠‌‌‌‌‌‌‌⁠​⁠‌⁠‌​‍‍​‍​‍‌⁠⁠‌

I’ve had fewer timeline surprises since I run designs through an AI scorer and lock the rule “no exonic off‑targets in the top‑10 hits” before any orders go out. It keeps the edit set and risk notes aligned for grants, though it can filter out some otherwise tempting guides. If you need a free option, CRISPOR’s batch view has been fine for me: https://crispor.tefor.net.

‌⁠‍⁠​‍​‍‌⁠‌​​‍​‍​⁠‍‍​‍​‍‌‍‌⁠‌‍​‍‌‍‍‍​⁠​⁠​‍​‍​‍⁠​​‍​‍‌‍‍⁠​‍​‍​⁠‍‍​‍​‍‌⁠​‍‌‍‌‌‌⁠​​‌‍⁠​‌⁠‍‌​‍​‍​‍⁠​​‍​‍‌‍‍‌‌‍‌​​‍​‍​⁠‍‍​⁠‌‍​⁠‍​​⁠‍​​⁠‍​​‍⁠​​‍​‍‌‍‌​​‍​‍​⁠‍‍​‍​‍​⁠​‍​⁠​​​⁠​‍​⁠‌‍​⁠​​​⁠​‌​⁠​‌​⁠‍​​‍​‍​‍⁠​​‍​‍‌‍‍​​‍​‍​⁠‍‍​‍​‍‌‍​⁠‌‌​‌‌‌‌⁠​⁠​​‌‍‌⁠‌​‍‍‌‌‍‍‌‍‌‍‌⁠​‍‌​‍​‌‍⁠⁠‌​⁠​‌​‍‍‌‍⁠‍‌⁠‍​‌‍‍‍​‍​‍‌⁠⁠‌