Mohammad Lotfollahi

@MohammadLotfol1

PhD candidate in Technical University of Munich (TUM). ML for biology.

Munich, Bavaria
Vrijeme pridruživanja: svibanj 2018.

Tweetovi

Blokirali ste korisnika/cu @MohammadLotfol1

Jeste li sigurni da želite vidjeti te tweetove? Time nećete deblokirati korisnika/cu @MohammadLotfol1

  1. proslijedio/la je Tweet
    29. lis 2019.

    Excited to share our scVelo manuscript - led by and , we generalize the beautiful RNA velocity concept from and to transient cell states through dynamical modeling. and

    Poništi
  2. proslijedio/la je Tweet
    8. lis 2019.

    Enjoyed reading this work by , & . Conditional VAE + improved regularization makes for very crisp interpretation of perturbations in scRNA-Seq (and beautiful batch-correction in the latent space).

    Prikaži ovu nit
    Poništi
  3. proslijedio/la je Tweet
    7. lis 2019.

    Excited to share our recent work on generative modeling for unpaired data using a ‚transformer VAE’. Led by and , we extend scGen using a conditional VAE together with an MMD regularization. Applications for images and scRNA-seq.

    Poništi
  4. proslijedio/la je Tweet
    7. lis 2019.

    Our (with , , ) new end-to-end-trained backend for scGen is out: transformer VAE. Predict how gut cells respond to infection, immune cells to stimulation, or how people would look with a smile (😐 -> 🙂).

    Poništi
  5. proslijedio/la je Tweet
    2. ruj 2019.

    Our project on extending models inspired by differential expression analysis to paired observations encountered in MPRA is out in ! allows you to perform DE-like worflows with sensible noise models on this data!

    Poništi
  6. proslijedio/la je Tweet
    22. kol 2019.

    Our pre-print "Deep learning at base-resolution reveals motif syntax of the cis-regulatory code" is out! Try training and interpreting BPNet on your own genomic tracks . Thx

    Poništi
  7. proslijedio/la je Tweet
    14. kol 2019.

    Our pre-print on predicting T-cell specificity to antigens based on TCR sequences is out: We built models on the new single-cell pMHC + TCR reconstruction data!

    Poništi
  8. proslijedio/la je Tweet
    4. kol 2019.

    New tool: scGen scGen is a generative model to predict single-cell perturbation response across cell types, studies and species

    Poništi
  9. proslijedio/la je Tweet

    A generative deep learning model that leverages ideas from image, sequence and language processing and applies these ideas to model the behaviour of a cell in silico is described in .

    Poništi
  10. proslijedio/la je Tweet
    1. kol 2019.

    Scientists at @HelmholtzMucEn have developed an for predicting a cell’s behavior in silico. It promises to reshape the way we study disease & on a cellular level.

    Poništi
  11. proslijedio/la je Tweet

    INSANE! For those at the first SCRUM, here is scGen, the tool that predicts single-cell perturbation responses!

    Poništi
  12. and in the end I have to thank my awesome supervisors and for their supports.

    Prikaži ovu nit
    Poništi
  13. as a side result, we show the model can also be used to correct batch effects in single cell datasets example: famous pancreases data set and also on MCA:

    Prikaži ovu nit
    Poništi
  14. and finally we show how one can use the model to predict across species:

    Prikaži ovu nit
    Poništi
  15. Next, we show our model can provide good predictions on disease (Figure3) and how it can transfer information from one study to another (figure4):

    Prikaži ovu nit
    Poništi
  16. In figure 2 we show how the model can capture effects of INF-beta on a PBMC dataset and also how it provides better performance compared to other methods. We further show model can capture cell type specific effects (Figure2 and S.F 5 and 6)

    Prikaži ovu nit
    Poništi
  17. In theory, scGen can use measurements of the effects of a stimulus on gene expression in one biological context to predict what would happen to the transcriptome if that stimulus were applied in a different biological context.

    Prikaži ovu nit
    Poništi
  18. My paper from first year of PhD is now out in (). We show a new algorithm that effectively "lifts over" differential expression patterns from a training sc-RNA-seq dataset to another, out-of-sample test dataset.

    Prikaži ovu nit
    Poništi
  19. proslijedio/la je Tweet
    29. srp 2019.
    Poništi
  20. proslijedio/la je Tweet

    An extremely smart use of latent space operations here! I wonder what more complex architectures would be able to achieve in the future, once adapted to biological context... fascinating

    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:

    ·