Assistant Prof. @ Stanford BASE, Genetics & Computer Science (courtesy). Lead the predictive biology lab of ML & single cell/spatial genomics, focus on heart
I am so proud of two talented RAs from my lab who helped transform our wet lab space to be fully functional in just three months! Yes! We will also do wetlab work too. Specifically, we will implement, optimize and develop new single cell, multi-omics and spatial transcriptomics
I am profoundly humbled and excited to officially announce that I will start my independent lab at Stanford this Dec. 16th. Growing up as the son of farmers in a remote village in China and facing the loss of my father at an age of 10, this journey has been anything but easy.
After 3.5+ years of perseverance, our dynamo paper is finally online in Cell today! Dynamo is an innovative and powerful framework of reconstructing continuous vector fields for mechanistic and predictive modeling of single cell fate transitions. Link:
Excited to announce that I officially started my predictive biology lab of single cell/spatial genomics at Stanford! Read more about our research here: Hiring at all levels (dry & wet). Stanford grad & undergrads are highly encouraged to work with us
After >2000 commits, I am thrilled to announce the release of dynamo 1.0.0! In this release, we pushed the limits of our continuous vector field approaches to develop an awesome set of techniques to enable functional and predictive analyses of (time-resolved)scRNA-seq experiments
Wonder how to take RNA velocity to the next level with labeling based scRNA-seq? Check our new tutorials () on the time-resolved vectorfield analysis for cell cycle/intestine organoids of scEU-seq
@AlexandervanOu1
and neural activation of scNT-seq
@hao_wu_7
🔬My sc/spatial genomics lab (Launch at
#Stanford
on Dec. 16) is seeking experimental RAs & post-docs in genomics & dev bio. Come to work with me and a vibrant team of sys biologists, ML experts, etc. to push the boundaries of gene regulation of development & disease.
#HiringNow
After 1200 + commits, I am proud to announce that dynamo is now on PyPi. See how we leverage it to go beyond RNA velocity to velocity / acceleration vector fields which then power us to calculate the potential, curl, divergence, curvature of single cells.
Thrilled to share that Dynamo v1.4 () for single cell and metabolic labeling enabled RNA velocity vector field learning and prediction is now on PyPi! This is one of our biggest upgrades of Dynamo since early 2023. In this new release, we made the following
Thrilled to share spateo-viewer: the “Google earth” App of spatial transcriptomics (ST) that allows interactive 3D alignment of ST, dynamic data exploration, etc. Kudos to Jiajun
@JiajunYao14
from Spateo development team on this beautiful work! More at:
Can we predict the single-cell transcriptomic state backward or forward over arbitrary time-scales (under certain reasonable assumptions)? We developed dynamo and show this is plausible by predicting long-term evolution of single cell with scSLAM-seq data.
With the imaging-based spatial transcriptomics such as MERFISH, seqFISH, CosMx SMI, Xenium and others, have you ever wondered how we can leverage their subcellular spatial information? Check out our latest preprint on Focus by Qiaolin and Jiayuan
@JiayuanDing
, two talent
We are excited to release dynamo 1.1.0. In this release, we highlight the in silico Perturb-seq and optimal reprogramming path predictions with a few tutorials, and added 50+ new functions. Please try these cool predictive analyses out and let us know how everything goes.
Excited to share Storm () for transient dynamics in gene-cell specific kinetics & RNA velocity via metabolic labeling augmented scRNA-seq. Kudos to Qiangwei from Tiejun Li's lab! Storm extended dynamo & will be incorporated in v1.4:
This is really overdue, but so glad that my last PhD work, Scribe (), was finally out. The finest work of my PhD career. Blessed to have Cole as the best advisor
@coletrapnell
. & very much thanks to
@sreeramkannan
and
@ARahimzamani
for the amazing colllabo
Many colleagues asked me what makes dynamo innovative and powerful? The capability to make predictions of optimal reprogramming paths and in silico perturbation really make it unique. now we release the primers and tutorials of those predictions:
🚀 Excited to announce Dynamo v1.3 for advanced modeling of scRNA-seq and vector field learning & prediction! A big step towards our v2.0.0 goal by the end of 23'. Get it now on PyPi ()!
#RNAVelocity
. See fully refactored preprocessing module, more below
If you're interested in research assistant, PhD, or post-doc opportunities, I invite you to reach out. Let's explore how we can work together or collaborate! email: xqiu
@wi
.mit.edu (or xiaojie
@stanford
.edu)
I am honored to be selected as Arc's inaugural Ignite Award recipient, among an amazing pool of colleagues. Really looking forward to joining the extended Arc family and interacting/collaborating with its core/innovation investigators & other recipients.
My lab will focus on the intersection of systems biology, machine learning, single cell, and spatial genomics. Our aim is to develop predictive models for understanding development and disease, with a particular emphasis on the heart.
My perseverance and dedication have led me towards this notable achievement. I can only hope that my family shares in this pride, knowing every step was in honor of our shared dreams and their incredible supports over the years.
We presented preliminary results of a continuous vector field approach, dynamo, a while ago. Now I am proud to tell you that we have established the full-fledged theory, and proved its broad utility, including app to COVID-19. See a full new version here:
Congrats, Felix! Mind-blowing work. This in principle solved the a key problem in single cell genomics that doesn't allow longitudinal observation of the same cell time. A much elegant solution than the LIVE-seq approach from nature recently!
Want to sequence RNA from cells without killing them or deliver RNA programs from cell to cell? Today in Cell, we present RNA exporters, which package and secrete cellular RNA, for non-destructively monitoring cell dynamics by sequencing and delivering RNA
How to make sense for a single-cell metabolic labeling experiment (scNT-seq, sci-fate, NASC-seq, scSLAM-seq)? check our newly released tutorials for dynamo (): . Many more to come soon!
Finally! We now provide detailed Dynamo tutorials for two metabolic labeling datasets and six other "conventional" scRNA-seq datasets. Check those beautiful velocity field plots and learn more about the tutorials and get the newest Dynamo here:
Curious about how to generate an awesome animation like the one for the cool scNT-seq neural activation dataset () with dynamo? Now we updated our tutorial to show how to produce animation in a just a few lines of code here:
Excited to share you that since today dynamo () has its own logo! Here the arrow represents the RNA velocity vector field, while the helix the RNA molecule and the colored dots RNA metabolic labels (4sU labeling).
congrats! I am pleased to be one of reviewers for this simple but interesting approach to improve splicing RNA velocity estimation: it fits a NN for each gene with a cosine correlation loss which leads to gene-wise velocity nicely flows the curvature of the data in a phase plot
My brilliant collaborator Yan
@YanZhan92480937
will talk about exciting updates of dynamo at . Dynamo now is powered by a principled approach for constructing a Markov transition matrix via "Itô kernel". Come to talk with him if u happen to be in the conf.
We will be a part of the Stanford BASE program
@StanfordBASE
, jointly with the Department of Genetics
@StanfordMed
, and affiliated with the Department of Computer Science
@StanfordEng
.
Scribe recovers CAUSAL networks and visualizes direct/in-direct causal regulations and combinatorial regulations. We show that it works the best on dynamics-coupled single-cell time-series measurements (RNA-velocity or potentially other alternatives)
Have some 10x, cite-seq or single-cell version of SLAM-seq data and want to play with the RNA/protein velocity, learn transcriptomic vector fields, map potential landscape? Check out our dynamo cheatsheet! Dynamo is fully functional and in beta stage now:
vector field reconstructed can then help to make animations to reveal RNA velocity/speed/acceleration/fate commitment in real-time!! In addition to the above animation, you can also check the animation for neural activation ()
We recently reported an update () on moving toward differentiable RNA velocity vector fields. To make the innovation in the work accessible to dynamo users, I will present a series of posts, starting from dynamical systems + differential geometry analyses.
My gratitude extends to all my previous/current funding supports such as the Jameel Clinic
@AIHealthMIT
, Impetus
@impetusgrants
, CZI
@czi
, K99
@NIH
, and others. Thank you all for believing in my work and vision.
Single cell SLAM-seq methods, similar to RNA velocity, measure the SAME cell TWICE and thus will improve causality detection. Check our new notebooks on visualizing/inferring causal network with scSLAM-seq, NASC-seq datasets, powered by Scribe .
Amazing work from Wang lab at Broad on time-resolved, and spatially-resolved transcriptomics via TEMPOmap. See how they can quantify RNA kinetics over time in subcellular resolution for thousand of genes. Great honor to contribute to this amazing work.
Excited to share TEMPOmap, our new spatiotemporal transcriptomics method that integrates metabolic labeling with 3D in situ sequencing to simultaneously profile the age and location of individual RNA molecules led by
@JingyiRen
Haowen Zhou and
@huzengHZ
Why single-cell SLAM-seq datasets give dramatically improved RNA velocity estimations (Erhard et al. 2019)? We released a jupyter notebook powered by dynamo () to show underlying analytical reasons and demonstrate with an analysis of (Hendriks et al. 2019)
Thanks
@tsuname
! Really unimaginable until very recently - now we can get those geometric quantities thanks to vector-field functions learnt from data with dynamo. Acceleration/curvature can be defined for each gene, you can use them just like expression or velocity matrices.
I would be remiss not to mention the invaluable support, guidance, and mentorship I received from my post-doc advisor
@JswLab
my PhD advisor
@coletrapnell
, and a host of incredible collaborators over the years.
Really enjoyed the collaboration with the amazing
@hao_wu_7
@QiQiu8788
@PengHu_xqq
! Also very proud to show dynamo can take full advantage of scNT-seq to recover "time-resolved RNA velocity" of neural activation while conventional scRNA-seq cannot!
Hey Scribe users, I just noticed a long-living but serious bug in causal_net_dynamics_coupling that leads to wrong calculation of causality Score (see equation on the left and the fix below). This has been fixed. I apologize for this and please update the analysis if you used it!
congrats
@WangXiaoLab
@JingyiRen
@hzhou99
. Honored to contribute to this amazing work to track the RNA production, nuclear exportation, trafficking and degradation at subcellular resolution with the TEMPOmap.
Congrats to
@JingyiRen
,
@hzhou99
, and the great team! With TEMPOmap, we can snapshot RNA life cycle from birth to death for thousands of genes with single-cell and subcellular resolutions, revealing that RNA kinetics, fast and slow, serves gene functions.
Importantly, we show that a comprehensive modeling framework of the time-resolved metabolic labeling data can be used to overcome some fundamental challenges of conventional splicing based RNA velocity analyses, which also give us absolute kinetic parameters / velocity.
Super cool work from
@jmartinrufino
in Cell to massive parallel creating genetic variants across genes and coupled w/ scRNA-seq to reveal downstream effects w/ many clinical potential. Delighted to contributed to the downstream velocity analyses w/ dynamo
Thrilled to share my PhD work in
@CellCellPress
! Massively parallel base editing to map variant effects in human hematopoiesis. We developed screens on blood stem cells to understand disease and develop treatments at single-nucleotide resolution.1/n
Lastly, with the completion of this work, it is a good time for me to transition to the next phase. I am seeking the opportunity to lead a team to apply genomics advances and to build novel models to understand the cell fate, and to predict and manipulate it.\
The key novelty of dynamo lies in its ability to take discrete velocity vectors to learn continuous functions of vector fields in transcriptomics space with which we can derive, analytically, RNA Jacobian, acceleration, divergence, etc, to perform differential geometry analyses
scNT-seq, sci-fate, scEU-seq, NASC-seq and scSLAM-seq are all metabolic labeling based scRNA-seq technologies. You may be also interested to read our new introduction page on summarizing those technologies and discussing how dynamo models them:
btw, with the contribution of this work and my previous dynamo paper, it is a good time for me to transition to the next phase. I am seeking the faculty job to lead a team to apply genomics advances & build novel models to understand the cell fate, and to predict & manipulate it
This work provides the foundation for my recent preprint of dynamo. I am working actively to seamlessly integrate Scribe () w/ dynamo () so that we can infer accurate causal regulatory networks from an improved RNA velocity framework.
@slinnarsson
@NimwegenLab
the RNA-velocity paper is definitively one of the most exciting paper I ever read in the single cell genomics field. It is based on a simple model but the results are surprising great. I am shocked that it is rejected.. best wishes for the appeal
Furthermore, we can perform non-trivial predictions that accurately predict the optimal paths and transcription factors that enable cell fate conversions, as well as the genetic response and cell fate diversion after the genetic perturbations.
scNT-seq is arguable one of the most scalable/ accurate tech for metabolic labeling scRNA-seq. I believe together with sci-fate/NASC-seq/scEU-seq/scSLAM-seq and potential others, the single cell field will witness a paradigm shift in measurement/analysis of single cell genomics
Jameel Clinic PI Jonathan Weissman (
@JswLab
), postdoc
@Xiaojie_Qiu
, and collaborators at the University of Pittsburgh School of Medicine are working on a ML framework called "dynamo" that will help predict cell fates and genetic perturbations.
Read more:
If you are interested in metabolic labeling enabled or regular scRNA-seq RNA velocity field, spatial transcriptomics, and quantitative modeling of single cell genomics in general, feel free to reach out to me. We are continuing developing dynast, dynamo, spateo and many more
It feels bittersweet to finish my last week in
@davidasinclair
lab. I was so lucky to find this dream match in the 4th rotation and start an incredible 6y journey with some low lows and high highs🍾. Now onto the next chapter with
@JswLab
to keep learning and growing💪. Excited!
If you are a PhD/Master student at Stanford, please come to talk with us on many exciting wetlab as well as dry opportunities (See opening positions in our website here: ). Job post for post-doc positions is copied below as well:
This tutorial is long due (the features are already in dynamo > half a year already) and there are many new functionalities in this release that are not demonstrated in any tutorials. Be sure to go back to check for more that are currently in pipeline. any feedbacks are welcome!!
My great thanks also go to
@shayanhoss
from
@spyros_darmanis
lab who contributed tremendously to help generate the clonally-traced scSLAM-seq/NASC-seq dataset used for the initial submission.
I moved with Weissman lab to WI, MIT recently and am excited to continue my research there. If you are in Boston area, interested in the concepts introduced in dynamo, please let me know. Happy to chat and look forward to some collaboration, especially the experimental side.
The Scribe manuscript is finally out on biorxiv after an epic delay! very proud of it, one of my best in my PhD. We show the importance of dynamics time coupling for causality detection. Hope this message will lead to a new era of research. Great collaboration with
@sreeramkannan
⏩ For the next release Dynamo v1.4, we'll bring new methods for metabolic labeling enabled, time-resolved
#scRNAseq
(e.g. scSLAM-seq, NASC-seq, scEU-seq, sci-fate, scNT-seq) + more enhancements on vector field learning. Stay tuned!
#bioinformatics
#RNAseq
Thanks
@tsuname
for speaking very highly of the conceptual innovation of our work! Bridging small-scale systems-biology/physics type of thinking with high-dimensional genomics using ML is my long-time dream. We are still pushing the limits of our continuous vector field approach
New post on on a convergence in how we model and conceptualize increasingly dynamical systems in biology and software.
Cells, neurons, molecules, machine learning, causality, polygenic models all make an appearance.
p.s. I also want to thank the editor, all 3 reviewers. Their constructive comments encourage us to profile the hematopoiesis which strengthened our work. We are also encouraged to push the limits of dynamo to make non-trivial predictions of optimal reprogramming paths, etc...
Spateo is our latest efforts to build Aristotle ecosystem, a novel full-stack computational ecosystem that provides advanced spatiotemporal modeling of single cell and spatial genomics datasets. This work is supported by
@cziscience
() and
@impetusgrants
Congrats on this amazing work on revealing the single-cell clinical COVID19 neutrophil response and dexamethasone treatment kinetics
@sarthak_sin
@jeffbiernaskie
. Nice application of the analytical vector field approach from dynamo ()
We also marry the scalability of machine learning with the interpretability of dynamical systems to gain mechanistic and predictive insights. We reveal the minimal network governing Meg lineage’ early appearance and accurately predict TFs of hematopoietic fate transitions.
I would greatly appreciate it if you/your department can communicate with me in this regard. I welcome any valuable advice also. And of course, always welcome any comments or questions about our work and dynamo. Happy to collaborate in any form too!
Come to work in the Lander lab at the Broad. An amazing opportunity for recent undergraduate graduates who want to work on some exciting technology development projects before applying PhD. Contact
@jmartinrufino
for more details
I am extremely thankful for
@JswLab
incredible mentoring and support over the past few years. This work is not possible without
@YanZhan92480937
(co-first author) from
@jhxing001
lab who contributed enormously to the theoretical undertaking of this work.
Congrats Liangcai
@GuLiangcai
, Xiaonan, Li and Runze and all others for this amazing work! PIXEL-seq with polony gel stamping that dramatically reduce time andcost will be another game changes in single-cell or subcellular resolution spatial transcriptomics!
@benoitbruneau
@BGI_Research
@MLongqi
currently stereo-seq can be accessed via research collaboration with BGI (you can directly email Longqi: liulongqi
@genomics
.cn). There are already kits for it and plans for commercializing it soon. it should be able to sequence the library with illumina too but
So implied in the logo, dynamo utlizes scRNA-seq, especially metabolic labeling enables ones, to learn the functions of vector fields to make non-trivial mechanistic and predictive insights: . Treat thanks to
@dummyIndex
and the team for this elegant design
We first introduce key concepts in dynamical systems and differential geometry to single cell genomics: e.g. topology of vector fields predicts stable/bifurcating phenotype while Jacobian state-dependent regulation. We also extend RNA velocity to acceleration/curvature fields
As a demonstration, we segmented single cells for an entire E16.5 embryonic slice and treat these data as conventional single cell RNA-seq to perform clustering. We then project cells back to physical space to reveal the intricate spatial distribution of different cell types.
As always, we welcome any comments & suggestions on our work. As I am starting my lab at Stanford & actively looking for talented research assistant/post-doc, etc., if you are interested in these kind of research, please email me at xqiu
@mit
.edu. Excited to working with you!