All materials from our Machine Learning for Single Cell Analysis workshop (January 2021) are available freely online now, including the lectures. Everything can be found at (including a link to the lectures on youtube ) !
Happy to share that the board of permanent officers both of
@YaleMed
and
@Yale
FAS have approved my tenure starting today. Honored to be one of the very few people tenured by both units. Glad to be a fully joint associate professor in
@YaleCompsci
and
@YaleGenetics
!!
Comp bio folks: can we all start to put our method in the "Results" section? I am reviewing a paper whose contribution is a beautiful method ... that you have to excavate from supplementary information. My first Result section in every paper I wrote is the method. It works!
(1/10)Pleased to share that our work on leveraging deep transformer-based representation learning for protein sequence optimization is out today in
@NatMachIntell
!
Honored to receive the NSF CAREER Award for our proposal: "Deep representation learning for exploration and inference in biomedical data." Thanks to the entire lab for the great work!
Congrats to Smita Krishnaswamy for winning the Excellence in Science Award Early-Career Investigator Award from the FASB! The award recognizes the research, teaching and community efforts she has made in the scientific community:
Some perspective for scientists who may be asked to review computational methods: Computational methods are indeed innovated by taking existing mathematical and algorithmic atoms and putting them together in a novel way. Second, small changes in the steps can have a large…
(1/n) We are excited to present a new multiscale data exploration tool called CATCH to identify signatures of immune activation in the first single cell atlas of human retinas affected by age-related macular degeneration (AMD)
Our lab, along with
@YaleGenetics
@YaleSEAS
is excited to bring you the third edition of our Machine Learning for Single Cell Analysis Workshop, totally virtual in January!
Yale is starting a Section of Biomedical Informatics and Data Science, looking for a Chair of the section who will also be Associate Dean for Bioinformatics at Yale School of Medicine (YSM). Apply here!
2023 was a great year for the
@KrishnaswamyLab
with papers in Cell, Nature Computational Science, Nature Communications, NeurIPS, ICML, LOG, IEEE MLSP, SIMODS, and many more, released many methods like TRELLIS, Temporal PHATE, Heat geodesic embeddings, Neural FIM, RITINI,…
Reviewers if you read a new method in a bio paper, please don’t dismiss it because package X has some related method Y. Y being in X does NOT make Y: 1. a gold standard, 2. unimprovable, 3. vetted, 4. benchmarked, 5. or even peer reviewed! Encourage innovation, we need it!
(1/n) Excited to present temporal-PHATE, lead by
@ericalbusch
, with
@mrguywolf
@g_lajoie_
@JcsHuang
and Andrew Benz, a manifold learning approach for time-series data. We apply T-PHATE to fMRI data to unveil cognitively-meaningful brain dynamics
How do you embed DATA^2 where each datapoint is itself a whole dataset?! Think massively multiplexed experiments... Read our work by
@wschen_
,
@mrguywolf
,
@david_van_dijk
together with
@BodenmillerLab
on our new embedding PHEMD out today in Nat Methods:
@emgbar
I feel you. I have this kind of thing happen to me all the time. When people ask me what I do, I say "professor" and as close as 2 miles from campus they say "where?" Because it can't possibly be Yale right :)?
We are proud of
@DBBurkhardt
who is defending his thesis tomorrow! It has been amazing to watch him grow as a computational scientist and person over the past 5.5 years!
(1/n) Excited to present our next advance in inferring dynamics from static snapshot data w/ dynamic optimal transport and neural ODEs: MIOFlow at
@NeurIPSConf
as a spotlight talk! This work was lead by
@Guillaume_hu
, Sumner Magruder,
@AlexanderTong7
!
Looking for postdocs with strong computational and mathematical skills, who are interested in quantitative problems inspired by big biomedical data. Experience in machine learning or signal processing is a bonus!
Happy Diwali from
@KrishnaswamyLab
! Celebrations at
@Yale
are strong. Today is the second day, Balipadyami, celebrating the legend of the Demon King Bali's ultimate generosity and goodness (Bad guy gone good!).
Want to learn some ML during your time at home? We are pleased to partner with
@YaleGenetics
to offer our workshop on Machine Learning for Single Cell Analysis ONLINE in May ! Registration is now open!
Excited to bring together trainees and scientists from everywhere for our Machine Learning for Single Cell Analysis workshop, starting tomorrow! Nice job
@DBBurkhardt
@scottgigante
and all others involved!
@AlexanderTong7
and I are happy to present Diffusion EMD, a new graph-based formulation of EMD
@icmlconf
2021 ()! This is an extremely fast, differentiable way of computing EMD between large sets of high dimensional distributions. Keep reading for more..
We are going to
@NeurIPSConf
! Come see us at
to learn more about a unifying framework for non-linear dimensionality reduction using heat diffusion. Or DM/email me if you are there and want to chat!!
(1/n) Excited to present collaborative work with
@VirusesImmunity
@fperrywilson
, Yale Impact,
@Mila_Quebec
,
@mrguywolf
! We explore 54 million cells, 18 clinical variables from YNHH patients with Multiscale PHATE to find features predictive of outcome.
Guru poornima is an Indian holiday celebrating all Gurus or teachers. I think the impact of teachers and advisors is particularly significant in academia. Take a moment to reflect and feel gratitude towards all those who are getting you through this journey! Namaste.
Happy Vijayadashami (Day of Victory) to all! I'd like to think of all the deadlines, the pandemic, the bugs in our code, the reviewers who don't get it ... as 10-headed monsters (Ravana) and our resilient, persistent students, and postdocs as the warriors (Rama, Lakshmana).
We are hiring! Looking for a postdoc who is interested in both neural networks and dynamic systems, with a strong mathematical or computational background. Multiple funded projects in ML for bio are open. Please email me your CV if you are interested.
Check out our new fast PHATE 2.0 for visualizing high-dimensional data. It features several improvements including faster runtime than tSNE.
#CZI
#HCA
#czihcacomptools
#SingleCell
In
@Patterns_CP
today ! Our single-cell Multimodal GAN (scMMGAN): a multiway GAN with a data geometric prior to align multiple data modalities from unpaired measurements and generate missing modalities! Lead by
@AmodioMatt
.
(1/n) Happy to share our work Time-Inhomogeneous Diffusion Geometry and Topology, published in the SIAM journal on the Mathematics of Data Science (SIMODS) !
At
@NeurIPSConf
2022, we’re introducing a novel scalar curvature for high dimensional data analysis, lead by
@dbhaskar92
and others, called Diffusion Curvature
(1/10)
Only on archive so far, but couldn't wait to share excitement about TrajectoryNet, by
@AlexanderTong7
@mrguywolf
@david_van_dijk
, a neural ODE framework that learns continuous dynamics from snapshot timecourse population data!
(1/n) Excited to share our first venture into patient data---exploring dynamic risk prediction using deep learning for patients with gastrointestinal bleeding
@SciReports
lead by (real) doctor and deep learner
@dlshung
!
Another preprint from our lab by
@david_van_dijk
and
@DBBurkhardt
on using neural networks to find a latent space that has natural archetypes (AAnet). We apply AAnet to TILs and microbiota, inspired by the work of
@UriAlonWeizmann
. Arxiv link here
Today:
#ICLR2022
workshop on Geometric and Topological Representation Learning! My foundation talk on Diffusion geometry at 10am EST to give you background. Then see our poster on Diffusion Curvature: a new way of measuring curvature in data (presented by
@dbhaskar92
) !
July 28 at
#ICML2023
@TAGinDS
"Heat Geodesic Perspective on Manifold-Preserving Dimensionality reduction," presented by
@Guillaume_hu
! We connect existing methods PHATE, diffusion maps and tSNE to heat diffusion on manifold (Varadhan's formula, Harnack's inequality), and…
If you like PHATE and neural networks, check out m-PHATE (with
@scottgigante
@gmishne
and Adam Charles) for visualizing network behavior, to be presented at
#NeurIPS2019
! See video:
Happy to be giving my talk today (7/14) @ 2pm ET - "Deep Representation Learning for Problems in Biology" (). I will be previewing some new techniques we've been working on, including sneak preview of a new version of PHATE!
Congrats to
@DBBurkhardt
on winning the Carolyn Slayman Prize in
@YaleGenetics
! This prize is awarded yearly to recognize remarkable achievements of students based on their body of work, the impact of their findings, and their commitment to graduate education. We're so proud!!
Happy mother's day to all moms! As a female academic, I often remember my mom telling me to stand up for myself, to be proud of who I am, and not to be afraid of being different...when I was 6 or 7 years old. Thanks mom for lessons in grit (and so much else)!
My daughter interrupting me about 3 times during my
@abacbs
talk (which I enjoyed btw), to 1) ask for my phone, 2) ask to turn on anya and elsya 3) can't even remember...But people asked lots of questions at the end so I consider it a victory :) .
Looking forward to presenting our first paper at RECOMB on inferring metabolic states from transcriptomic data using geometric deep learning. Lead by
@SiddharthVish
and
@hollysteach
!
I think this idea is cool. However, I _really_ worry how people will interpret these plots at first glance. It looks like a developmental tree. But if you do this on scRNA-seq, _every cell_ will be represented by a _curve_ going up from 0 along the y axis.
This year we had an amazing student from Israel, Noam Rozenboim, who just accepted
@YaleCompsci
PhD offer to work with us on a fantastic mix of math, deep learning in
#cancer
biology. But alas, she succumbed to the same disease before we met her in person
#CancerResearch
.
Check out our poster at
#ISMBECCB2023
by
@dbhaskar92
, Garrek Chan,
@AmodioMatt
jointly with collaborators at
@Boehringer
on performing embedding and other knowledge graph tasks with Geometric Scattering and our Feature-Mapping GAN framework!
Fascinating discussions on manifolds, topology, geometric measure theory, dynamics, graphs and more at the “Exploratory workshop on Math, Theory and Cancer Biology,”
#Math4CancerBio
workshop at the
@theNCI
today. Thanks for organizing Raul Rabadan and Shmuel Weinberger.
Happy belated father’s day to all fathers out there. There is no way I would have made it as an academic with 3 kids without amazing amounts of help from mine!
PHATE is a dimensionality reduction algorithm designed for visualizing all kinds of data. Here we show 16K differentiating stem cells measured with scRNA-seq. Unlike tSNE and UMAP, PHATE doesn’t create “blobs” and instead preserves continuous structures in the data (2/10)
Honored to be selected as the Kingsley Fellow for
@YaleMed
recognizing "outstanding junior faculty research and future research potential." I hope to work hard to fulfill this potential.
Excited to announce: “Deep thoughts on geometric learning & exploration of non-Euclidean data” organized by
@mrguywolf
(
@MILAMontreal
,
@UMontreal
) and Matthew Hirn (
@CMSE_at_MSU
) as part of SIAM MDS. StartsJune 17th:
Check out our new method for enhancing experimental signals from single cell experiments with multiple samples measured under different experimental conditions: !!
good mom day:
6yr old daughter K: hey mommy what’s your job?
me: I’m a professor.
K: in what?
me: computer science and..
K: eee !!(squeeling) computer scientist!!
me: pretty cool right?
K: We talked about scientists in social studies class. I said that’s what I want to be!
(1/n) Excited to share Integrated Diffusion, a method for multimodal data integration that merges data views via diffusion operators for many downstream. Lead by
@manik
and
@abhinav
,
@AlexanderTong7
@mrguywolf
, published in IEEE MLSP!
Looking forward to the first day of our Machine Learning for Single Cell Analysis Workshop and meeting the students! Thanks to all the TAs and
@DBBurkhardt
and
@scottgigante
for the organization!
Congrats to Kincaid Macdonald on his senior thesis in Math
@Yale
on Diffusion Curvature and it’s relationship with coarse Ricci curvature. Kincaid has worked with us since freshman year, published 4 papers, TAed 3 classes, and more yet to come! We’re proud of you!
As a preview: The two papers to appear in
@NeurIPSConf
are with regards to learning dynamics on a manifold and estimating curvature on point cloud data !
(1/n) Check our FlowArtist method for embedding data with velocity vectors at
#MLSP2023
presented by
@dbhaskar92
--developed during Summer Undergraduate Math Research
@Yale
: Kincaid Macdonald, Guy Thampakkul, Nhi Nguyen, and
@ZhangJoia
, together Ian Adelstein and myself!
Happy to be at
@CSHL
single cell analysis for some Saturday science! Very engaged audience for our talks and great talks by
@CarolineUhler
and
@lpachter
!
Happy Thanksgiving! Let's all be thankful for our universities, brilliant colleagues, awesome students, meteoric postdocs and for the fact that we get up everyday and do what we love: science!
Congrats
@egbertcastro
or Dr. Castro as he is now known!! Passed his thesis defense with work on biomolecular representation. Can’t wait to see what you do next!!