Anita Donlic

Congratulations!! This is so exciting and well-deserved. 🎉

... Prof. Ai Ing Lim and @kristantunes.bsky.social
for a great help with the virus work, and Cliff @brangwynnelab.bsky.social for all the support and guidance!

To say that this took a village would be an understatement. I’m incredibly grateful to my collaborators in the @brangwynnelab.bsky.social : Troy Comi, @sofiquinodoz.bsky.social, @nima-jaberi.bsky.social, Lenny Weisner and Lifei Jiang, ...

🌟 Big picture 🌟
✔️ Condensate morphology encodes function
✔️ Deep-Phase translates images into quantitative readouts of drug biochemical potency
✔️ Future applications: phenotypic screens, MoA studies, biomarkers, single-cell heterogeneity detection (10/10)

Integrating our discoveries, we came up with a new model for nucleolar phase boundary maintenance, in which both the abundance and processing state of rRNA📊 dictate the multiphase arrangement within this fascinating condensate (9/10).

These observations led us to hypothesize that sequential cleavage inhibition💊 (invert & push apart DFCs from GC) followed by transcription inhibition💊 (restore DFC-GC interface) could phenocopy the flower 🌸 morphology. This is indeed what we (and Deep-Phase) saw! (8/10)

We then conduct mechanistic studies 🔎 to find that beyond its recognized role in resolving supercoiling, TOP1 is also involved in rRNA processing! Its inhibition causes rRNA cleavage defects and a drop in transcription of large ribosomal subunit precursors ❌🏭. (7/10)

To test the utility of Deep-Phase, we conduct a small molecule screen and uncover a unique nucleolar morphology: the "flower" 🌸! This phenotype, where the DFC-GC interface is maintained but inverted, arises specifically upon DNA Topoisomerase I (TOP1) depletion. (6/10)

Not just nucleoli! We extend Deep-Phase to nuclear splicing speckles✂️ as well as RSV viral cytoplasmic factories🦠, showing that condensate morphology alterations are a quantitative fingerprint of the degree of RNA biochemistry disruptions within them. (5/10)

In doing so, we measure dose-response curves from images and demonstrate that they match IC50 values from biochemical experiments: nucleolar morphology becomes a quantitative readout of potencies of drugs that alter ribosome biogenesis! 📉➡️📈 (4/10)

We developed Deep-Phase, an automated imaging and deep neural network-based framework that accurately tracks these changes over treatment times⏲️and concentrations📶 directly from images (no hand-picked features) (3/10).

Cells organize key processes in biomolecular condensates, such as the multiphase nucleolus where ribosomes are made 🏭. Due to its dynamic and compartmentalized nature, small molecule perturbations of this process💊 lead to fast and distinct morphology rearrangements🌀. (2/10)

Deep Learning of Functional Perturbations from Condensate Morphology Biomolecular condensates compartmentalize the interior of living cells to spatiotemporally organize complex functions, yet linking molecular interactions within condensates to their mesoscale organiza...

I’m thrilled to share our preprint that uses deep learning to interrogate structure-function relationships in condensates! biorxiv.org/content/10.1... In here, we ask: can AI read condensate biology from pictures alone? Turns out yes... see what we discover below! 🤖🧬🖼️ (1/10)

Ai Ing Lim and @kristantunes.bsky.social for a great help with the virus work, and Cliff @brangwynnelab.bsky.social for all the support and guidance!

To say that this took a village would be an understatement. I’m incredibly grateful to my collaborators in the @brangwynnelab.bsky.social : Troy Comi, @sofiquinodoz.bsky.social, @nima-jaberi.bsky.social , Lenny Wiesner and Lifei Jiang for their hard work and expertise, ...

🌟 Big picture 🌟
✔️ Condensate morphology encodes function
✔️ Deep-Phase translates images into quantitative readouts of drug biochemical potency
✔️ Future applications: phenotypic screens, MoA studies, biomarkers, single-cell heterogeneity detection
(10/10)

Integrating our discoveries, we came up with a new model for nucleolar phase boundary maintenance, in which both the abundance and processing state of rRNA📊 dictate the multiphase arrangement within this fascinating condensate. (9/10)

These observations led us to hypothesize that sequential cleavage inhibition💊 (invert & push apart DFCs from GC) followed by transcription inhibition💊 (restore DFC-GC interface) could phenocopy the flower 🌸 morphology. This is indeed what we (and Deep-Phase) saw! (8/10)

We then conduct mechanistic studies 🔎 to find that beyond its recognized role in resolving supercoiling, TOP1 is also involved in rRNA processing! Its inhibition causes rRNA cleavage defects and a drop in transcription of large ribosomal subunit precursors ❌🏭. (7/10)

To test the utility of Deep-Phase, we conduct a small molecule screen to uncover a unique nucleolar morphology: the "flower" 🌸! This phenotype, where the DFC-GC interface is maintained but inverted, arises specifically upon DNA Topoisomerase I (TOP1) depletion. (6/10)

Not just nucleoli! We extend Deep-Phase to nuclear splicing speckles✂️ as well as RSV viral cytoplasmic factories🦠, showing that condensate morphology alterations are a quantitative fingerprint of the degree of RNA biochemistry disruptions within them. (5/10)

In doing so, we measure dose-response curves from images and demonstrate that they match IC50 values from biochemical experiments: nucleolar morphology becomes a quantitative readout of potencies of drugs that alter ribosome biogenesis! 📉➡️📈 (4/10)

We developed Deep-Phase, an automated imaging and deep neural network-based framework that accurately tracks these changes over treatment times⏲️ and concentrations📶 directly from images (no hand-picked features).(3/10)

Cells organize key processes in biomolecular condensates, such as the multiphase nucleolus where ribosomes are made 🏭. Due to its dynamic and compartmentalized nature, small molecule perturbations of this process💊 lead to fast and distinct morphology rearrangements🌀. (2/10)