Shian Su

840 posts

Shian Su

Shian Su

@shian_su

Research Officer in Bioinformatics at the Ritchie Lab, WEHI. Likes #rstats, #nanopore, #datavis and mangos. Views are a great C++17 feature.

Melbourne, Victoria Katılım Aralık 2015
387 Takip Edilen316 Takipçiler
Shian Su retweetledi
Heng Li
Heng Li@lh3lh3·
Our review paper on "Genome assembly in the telomere-to-telomere era" published online in @NatureRevGenet. It amazes me how much assembly has progressed in the past four years – often ~100X improvement in contiguity and base accuracy and with haplotypes and hard regions resolved!
Nature Reviews Genetics@NatureRevGenet

Genome assembly in the telomere-to-telomere era go.nature.com/4aLqIED #Review by Heng Li @lh3lh3 & Richard Durbin @richard_durbin @DanaFarber @HarvardDBMI @Cambridge_Uni

English
3
125
411
46.5K
Shian Su retweetledi
max
max@Max_Deforet·
Stop the video and tell me if you can locate the disk (I bet you can't). When even your eyes cannot identify cells in a still image, improving the segmentation method is pointless. A better approach is to use temporal information from previous frames and next frames.
PRX Life@PRX_Life

This PRX Life paper debuts DisNet2D, a deep #NeuralNetwork for image analysis. The method merges tracking and segmentation to achieve low error rate, allowing advanced image analysis of densely-packed cells. 📝 go.aps.org/4430Ykx

English
5
37
154
36.3K
Shian Su
Shian Su@shian_su·
@strnr @i000 In my experience, very few solvable problems are truly that niche. My workflow is to start with a chat with GPT-4 to break down a problem. Then annotate intent with comments as I code with co-pilot.
English
0
0
0
61
Stephen Turner 🦋 @stephenturner.us
@i000 I already did! My point was that the useless suggestions I kept getting on what was such a niche bioinfo problem that couldn't have possibly been in the training set were more distracting than helpful.
English
1
0
0
42
Stephen Turner 🦋 @stephenturner.us
OTOH, I haven't asked or answered a SO post in >1 year. Copilot and GPT-4 are great for problems that have been solved before, writing tests, documentation. I've just found copilot distracting for niche problems that it couldn't possibly understand without thorough explanation.
Stephen Turner 🦋 @stephenturner.us@strnr

My most contrarian, controversial, and cancellable observation this year: I disabled Copilot in VSCode and RStudio yesterday and (subjectively) my creativity and productivity went *up*.

English
2
1
7
1.9K
Shian Su retweetledi
Prof. Heejung Chung
Prof. Heejung Chung@HeejungChung·
I've been waiting for this moment and it has finally happened. I got a paper review back saying I need to familiarise myself more with the works of Heejung Chung and that my work should engage more with her work.
GIF
English
278
2.9K
54.1K
1.8M
Shian Su retweetledi
Dr Harriet Dashnow
Dr Harriet Dashnow@hdashnow·
I am currently recruiting a computational postdoc in my lab at CU dashnowlab.org The candidate would develop and use computational genomics methods to understand the genetic underpinnings of rare diseases and increase diagnoses. Pay starts at $70k for recent PhD grads
English
0
34
65
14.4K
Shian Su
Shian Su@shian_su·
Any function named read_csv should never have a "delimiter" argument and this is a hill I will die on.
English
1
1
1
139
Shian Su retweetledi
Quentin Gouil
Quentin Gouil@QGouil·
Females are #epigenetic mosaics! as illustrated by 🐈. Knowing which X chromosome is active is essential to understand how X-linked disorders affect females. We show how this can be done efficiently with @nanopore long-read sequencing... doi.org/10.1101/2024.0…
Quentin Gouil tweet media
English
3
17
81
7.6K
Shian Su retweetledi
Bo Wang
Bo Wang@BoWang87·
(1/5) Attention, Cellular Imaging Enthusiasts! Thrilled to share that our Analysis "The Multi-modality Cell Segmentation Challenge: Towards Universal Solutions" has been published in Nature Methods (@NatureMethods) 🎉🎉🎉 Cell segmentation is a critical component of microscopy image analysis, but existing algorithms are often specialized for particular types of images or demand users to manually set hyper-parameters, posing challenges for biologists without a strong computational background. To address these issues, we organized an international challenge at NeurIPS 2022 to promote the development of novel cell segmentation methods that excel across a wide array of microscopy images, imaging platforms, and tissue types. 🌐 Homepage: uni-cellseg.github.io 📄 Paper: nature.com/articles/s4159… 🔬Data: neurips22-cellseg.grand-challenge.org/dataset/ 🏅Competition website: neurips22-cellseg.grand-challenge.org 👇 👇👇👇
GIF
English
28
257
1K
215.2K