RT @LaurieRumker: We are excited to share Genotype-Neighborhood Associations, GeNA, a new tool adapting our CNA framework to detect cell st….

8/ I also owe a heartfelt thank you to amazing co-authors @PillaiPriyaP, @PetrosBrittany, @NicoleLWelch, and @cambearon, in addition to my labmates in the @sabeti_lab - it’s been such a joy and privilege to work alongside you all!.

7/ I am immensely grateful to @HaydenMetsky and @PardisSabeti whose endless encouragement, enthusiasm, and guidance made this project possible. They have been the most incredible mentors, and have taught me so much throughout the course of this journey.

6/ More broadly, our results suggest that artificial sequences designed by generative algorithms could advance the performance of any nucleic acid based diagnostic technology, and may also benefit the design of many other types of CRISPR guide RNAs.

5/ Excitingly, MEAs took advantage of Cas13’s protospacer and mismatch preferences to design optimal guides, despite not having these preferences explicitly encoded, and even illuminated new guide design principles that expanded our understanding of Cas13’s activation mechanism.

4/ We benchmarked the performance of MEAs across 20+ design tasks in several pathogens, from dengue virus to P. falciparum. In experiments, MEA-designed guides enabled more sensitive detection of highly diverse pathogens and more specific discrimination of pathogen variants.

3/ We applied MEAs to two different objectives: 1) designing guides that achieve maximal activity across sequence variation, and 2) designing guides that can optimally differentiate between highly similar sequences (e.g., two viral lineages).

2/ We developed model-directed exploration algorithms (MEAs), which combine a predictive model of CRISPR-Cas13a guide activity with search algorithms (evolutionary/WGAN-AM) to explore a sequence landscape and generate optimal diagnostic guides for any specified set of targets.

1/ Generative ML algorithms have emerged as powerful tools for designing biological sequences. We wondered if such approaches could advance guide RNA design, particularly the design of diagnostic guide RNAs that must detect an extraordinary degree of pathogen sequence diversity.

Delighted to share this preprint describing work in the @sabeti_lab, where we built new algorithms to design optimal CRISPR guide RNAs for pathogen detection! 🧬🦠. A brief 🧵….

RT @VaishnaviP27: My first first-author paper is now @CellReports! We discovered that the Golgi scaffold protein MMD regulates lipid metabo….

RT @VaishnaviP27: Thrilled to share that my first first-author manuscript is now on bioRxiv! We deciphered a new function for an understudi….

RT @LorenWalensky: Tonight we welcomed the amazing 2022 entering class of @HarvardMITmdphd students. The future of compassionate patient ca….

RT @AriztiSanz: Excited to see our paper finally out at @natBME! 🤗. Check it out @ . 1/3.

RT @PillaiPriyaP: I was so lucky to join the ADAPT team 2 years ago to work on an amazing paper published in @NatureBiotech on detecting vi….

RT @haydenmetsky: We wondered if ML-guided design could improve viral diagnostics—and found it does enable better sensitivity than existing….

RT @NatureBiotech: Designing sensitive viral diagnostics with machine learning

RT @s_keshavjee: Paul has left us too soon. A champion for working in solidarity with those in need, a teacher and inspiration to so many,….

RT @NicoleLWelch: The mCARMEN Accelerated Article Preview is live on Nature Medicine!! (with data tracking the Omicron wave!) Check it out….

RT @cambearon: Detecting viruses is important, especially these days. For the past five years, I have worked to develop CRISPR-based techno….