David Kelley

@drklly

Using computers to understand the molecular logic of life.

Vrijeme pridruživanja: listopad 2013.

Tweetovi

Blokirali ste korisnika/cu @drklly

Jeste li sigurni da želite vidjeti te tweetove? Time nećete deblokirati korisnika/cu @drklly

  1. Prikvačeni tweet
    17. lis 2019.

    In new work in collab with and Katie Pollard, we developed conv net architectures to predict 3D genome contacts from DNA sequence. For a complex process, these models work surprisingly well!

    Prikaži ovu nit
    Poništi
  2. 15. sij

    And check out our web interface into the data:

    Prikaži ovu nit
    Poništi
  3. 15. sij

    For example, his cell type classifier scNym trained to transfer annotations from the CZI mouse cell atlas achieves better performance than other available methods for us .

    Prikaži ovu nit
    Poništi
  4. 15. sij

    's systematic questioning of the data and creative approaches really improved our understanding of how these tissues age.

    Prikaži ovu nit
    Poništi
  5. 15. sij

    Check out the final version in Genome Research of our single cell RNA-seq profiling of lung, kidney, and spleen in aging mice!

    Prikaži ovu nit
    Poništi
  6. proslijedio/la je Tweet
    30. pro 2019.

    This adorable book is a gem. It'll make you feel sane and feed your irrational craving for charts at the same time. Also, excellent supplementary material for all of your resolutioning needs. #2020 More: 🙌:

    Poništi
  7. 2. pro 2019.
    Poništi
  8. proslijedio/la je Tweet
    26. stu 2019.

    If you haven't already, 's "One Strange Rock" is mandatory viewing. It intros the history, fragility, and complexity of our planet, and what it means for us today. Plus, it's just gorgeous. 🌍 👏: et al.

    Poništi
  9. 20. stu 2019.

    On my second day, they slipped "alias ls='ls ..'" into my .bashrc file, confusing me for hours 🙃

    Prikaži ovu nit
    Poništi
  10. 19. stu 2019.

    My first day at Maryland, I landed at a desk surrounded by those three incredible scientists. What a lucky break!

    Prikaži ovu nit
    Poništi
  11. 15. stu 2019.

    We also showed that you can further improve your doublet labels by combining Solo with cell hashing, which works better than either approach alone.

    Prikaži ovu nit
    Poništi
  12. 15. stu 2019.

    Check out our new method Solo for filtering doublets in scRNA-seq: Using autoencoder embeddings of single cell gene expression profiles and a neural network classifier, we improved doublet identification accuracy.

    Prikaži ovu nit
    Poništi
  13. proslijedio/la je Tweet
    12. stu 2019.

    Confession: we have a huge crush on , a magazine that does a gorgeous job of bringing together and . Give your eyes and mind a treat. Past issues here: 🙌: 🎨: & Christoph Straub

    Poništi
  14. proslijedio/la je Tweet
    8. stu 2019.

    "This is not just an academic issue. People whose lineages are well represented in genomic databases will benefit the most from precision medicine approaches..." 🖋️: Charles Lee via #scienceinsociety#medicalaccess

    Prikaži ovu nit
    Poništi
  15. proslijedio/la je Tweet
    6. stu 2019.

    A 2002 UK study found that kids were better at identifying Pokemon characters than actual wildlife. For, that was an opportunity. Enter: Phylo, an educational trading card game ft. beautiful . If you can't beat 'em, join 'em. 🐢

    Poništi
  16. proslijedio/la je Tweet
    31. lis 2019.

    "First, it eats them from the inside; then, it turns them into zombies." 🧟 📽️ Enjoy a spooky and watch how this fungus takes control of its fruit fly hosts

    Poništi
  17. proslijedio/la je Tweet
    23. lis 2019.

    AI will always reflect the biases of society, which means improving the algorithms isn’t enough. We need to improve the source material.

    Poništi
  18. proslijedio/la je Tweet
    21. lis 2019.

    Like it or not, the most convincing arguments aren't always the most logical ones. This quick run down introduces a few of the factors that matter most. When's the last time someone changed your mind? 🎥: 🗣️:

    Poništi
  19. 17. lis 2019.

    For an alternative approach, also check out DeepC

    Prikaži ovu nit
    Poništi
  20. 17. lis 2019.

    Using the latest hardware and , we can run 1 Mbp sequences through the model. That opens the door to directly predicting distant regulatory interactions.

    Prikaži ovu nit
    Poništi
  21. 17. lis 2019.

    Making this work required two additional stages on top of the Basenji stack. First, transform 1d sequence vectors to 2d to allow regulatory elements to recognize each other. Then treat the 2d representation as an image and run 2d convolutions to flesh out the rest of the map.

    Prikaži ovu nit
    Poništi

Čini se da učitavanje traje već neko vrijeme.

Twitter je možda preopterećen ili ima kratkotrajnih poteškoća u radu. Pokušajte ponovno ili potražite dodatne informacije u odjeljku Status Twittera.

    Možda bi vam se svidjelo i ovo:

    ·