Paul G. Allen Professor in the Allen School of Computer Science & Engineering at the University of Washington
@uwcse
AI/ML researcher; Computational biologist
Became a professor "fully" this quarter. It felt different when I got emails through full-profs
@cs
for the first time :) My full prof resolution is to bravely try completely new things on doing science, teaching, advising Ph.D. students, and mentoring people at various levels.
Deeply honored to receive the 2024 International Society for Computational Biology (ISCB) Innovator Award! My heartfelt gratitude goes out to my students and collaborators for their contributions in harnessing the transformative power of AI to drive breakthroughs in biomedicine.
Machine Learning in Computational Biology (MLCB), Dec 13-14 2019, Vancouver
We are excited to be holding the 14th MLCB meeting. Between 2004 and 2017, MLCB was an official NeurIPS workshop. Given the growth and maturity of the field, this year MLCB will be an independent meeting.
I am tremendously proud of
@joejanizek
, my
@uw
MD/PhD student (who just obtained CS PhD
@uwcse
). His explainable AI work on cancer pharmacogenomics, co-supervised by Prof.
@naxerova
@MGHCSB
@harvardmed
and me, is now accepted for publication in Nature Biomedical Engineering! 1/3
Finally published! 🥳🎉 Explainable AI has revealed that hematopoietic differentiation is a crucial indicator for identifying anti-cancer drug synergies in acute myeloid leukemia.
Check out our latest paper leveraging generative AI to dig into the flaws in the reasoning process of 5 dermatology AI devices! Explainable AI is no longer a luxury!
Our paper was published this morning in Nature BME:
News & Views:
Excited to announce our "deep profiling" of 18 human cancers! Incorporating all available cancer expression data into DNNs was my first idea 14 years ago when joining UW. Honored to work with brilliant minds,
@weiqiu55
@uwcse
and
@naxerova
@harvardmed
!
Excited to share that after a long embargo period, I've become a Fellow of
@aimbe
today! Grateful for the opportunity to collaborate with incredible students and collaborators.
Explainable AI can predict your chance of all-cause mortality (or general health status) and explain to you why the prediction was made. Check out my incredible
@uwcse
PhD student
@weiqiu55
's work, recently published in Nature Communications Medicine!
Excited to see our
@NatureMedicine
paper out today led by
@ChanwooKim_
with my co-senior author
@suinleelab
and an amazing team! We used a dermatology foundation model to enable explainable and transparent AI - from auditing datasets to models.
Excited to announce that I'll be giving the Distinguished Keynote Speech as the 2024 ISCB Innovator Award Winner! …
Join me at
#ISMB2024
to explore the latest in computational biology:
#ismb
“With TreeExplainer, we aim to break out of the so-called black box and understand how
#MachineLearning
models arrive at their predictions," says
@uwcse
prof
@suinleelab
about her lab's recent
@NatMachIntell
paper.
Details:
Can we use RNA-seq data to select an effective probe gene set for spatial transcriptomics? We answer this question in our study (Cover et al.
@ianccovert
in collaboration with Uygar Sümbül
@AllenInstitute
)
It has been tremendously rewarding to work with outstanding
@uwcse
students,
@HughChen18
,
@ianccovert
, and
@scottlundberg
. Our paper that reviews and unifies 26 distinct algorithms to estimate Shapley values is just published in Nature MI!
My talk at the "XAI in Action" workshop will begin at 10:30 am, and the second talk at the "AI for Science" workshop is scheduled for 3:05 pm. Please come join me!
🎉 Join us at NeurIPS 2023 AI for Science Workshop on 12/16:
7 speakers on cutting-edge AI research across fields🧠
Future-focused panel with funding agencies 💼
Open Catalyst Challenge announcement 🏆
(4/4) Publication news:
I cannot miss this important paper of Ian Covert, recently accepted to the Journal of Machine Learning Research.
Anyone interested in developing or applying ML interpretability methods should read this paper.
Thank you ICLR'23 for accepting our paper (spotlight) on explaining a vision transformer model! Congratulations to my enormous
@uwcse
PhD students (co-first authors),
@iancovert
and
@ChanwooKim_
.
If you want to know what your ViT pays attention to...you might not want to use attention values! Shapley values can do this better, and now they can even do it efficiently. Check out our new paper (ICLR spotlight) 🧵⬇️
Excited to announce the inaugural AIMBA (AI Meets Biology of Aging) 2024 meeting on May 22, 2023, 9-2pm PT, organized by
@UW
@NathanShockCtrs
!
Featuring four keynote speakers: Profs
@mariabrbic
, Anne Brunet, Vadim Gladyshev,
@james_y_zou
.
Check it out!
An amazing collaborative work of two of my MSTP (MD/PhD) students, and Dr. Nathan White in UW Emergency Medicine, is published in Nature Biomedical Engineering now.
Gabe Erion
@gabeerion
, and Joseph Janizek
@joejanizek
, now computer science Ph.D.s: I am so proud of you!
A cost-aware AI framework facilitates the development of predictive AI models that optimize the trade-off between prediction performance and feature cost.
So thrilled to work with my incredible
@uwcse
Ph.D. student
@ianccovert
on our paper about dynamic feature selection that just got accepted to ICML'23! Congratulations
@ianccovert
@weiqiu55
, Ming, and Nayoon !! 🥳🎉🍾
Pandemic graduation ceremony in Lee Lab style! Congratulations Gabe Erion
@gabeerion
for being the first computer science PhD as a UW MSTP (MD/PhD) student! I feel so lucky to work with these incredible people.
Another publication news:
The work of Gabe Erion, Joseph Janizek
@joejanizek
, and Pascal Sturmfels (alphabetically ordered co-first), which introduces the "attribution prior" and "expected gradients", got accepted to Nature Machine Intelligence as well.
Finally published. Congrats
@ianccovert
and
@scottlundberg
! This paper explains the mechanisms of 26 popular model explanation methods within a simple unifying framework:
Congratulations to
@joejanizek
,
@NicasiaBW
, and
@ianccovert
for graduating from
@uwcse
today!
I deeply appreciate your decision to choose me as your Ph.D. advisor, trusting me along the way, and joining our lab's journey to advance the fields of biomedical sciences and AI/ML.
In our recent preprint, we have delved into the shortcomings within the reasoning process of dermatology AI devices. Our paper sheds light on these flaws and their implications. 1/2
Happy to share that our paper on leveraging foundation models to foster the explainability and transparency of medical AI has been published today in
@NatureMedicine
! Check it out here:
Check out the recent Nature BME
@natBME
's News & Views on our paper: ! It has been a tremendous honor to work with exceptional young people, including the lead author
@uwcse
@UWMSTP
student Alex DeGrave, and the co-senior author
@RoxanaDaneshjou
(Stanford).
Scott Lundberg defended today: "Explainable AI for Science and Medicine" I noticed that our NeurIPS paper (Dec 2017) got cited 210 times, and our Tree SHAP paper is under review in a high profile journal.
Scott is the one who defended. I don't understand why I'm so exhausted.
Using generative models to dissect the medical AI reasoning process - done by a wonderful team of the first author
@uwcse
@UW_MSTP
student, Alex DeGrave, my co-senior author
@RoxanaDaneshjou
, and co-authors,
@joejanizek
and Zhuo Ran Cai.
New paper just dropped!
What if you could audit medical image AI algorithms using generative models partnered with human experts? We dissected 5 dermatology AI algorithms and found that models relied on both reassuring and concerning clinical features!
Check out the incredible work on the robustness of explainable AI
#XAI
methods done by two outstanding co-first authors:
@uwcse
Ph.D. student
@chrislin97
and a recent
@uwcse
graduate Dr. Ian Covert
@ianccovert
.
Please never say to your scientist friend that the only reason they are doing well is that she is just selective in picking good students! I know that some people have issues with women scientists of color, but you do not need to expose your insecurity that way. Just don't do
Super rewarding to work with such brilliant young minds,
@efweinberger
,
@ianccovert
, and
@chrislin97
. Congratulations on your NeurIPS'23 accepted papers! 🥳
"Feature Selection in the Contrastive Analysis Setting"
"On the Robustness of Removal-Based Feature Attributions"
#NeurIPS
The 16th MLCB 2021 meeting will be held on 11/22-23, 9 am - 5 pm PT. Registration is free. There will be 3 keynote speeches, 16 oral presentations, 21 spotlights, an industry panel discussion, and a virtual poster session. Details can be found: .
We have an upcoming paper at ICLR 2023 on a new feature attribution method for explaining representations learned by unsupervised models!
This was joint work with the fantastic
@HughChen18
@ChanwooKim_
and my advisor
@suinleelab
. (1/n)
Engaging with
@twimlai
on the future of
#ExplainableAI
and its role in addressing intricate challenges in biology and medicine was truly enlightening!
Had a fantastic time presenting at the ICML Comp Bio Workshop. Thanks
@bidumit
and
@mariabrbic
for the chance!
Today we’re joined by
@suinleelab
from
@UW
to discuss her work at the intersection of Explainable AI, computational biology, & medicine, including her talk from the recent
#ICML2023
Workshop on Computational Biology!
🎧🎥Check out the full conversation at
I am tremendously proud of my
@UW_MSTP
@uwcse
student
@joejanizek
who developed an explainable AI approach for unsupervised gene expression modeling, which led to an exciting collaboration with
@mkaeberlein
's lab on understanding the Alzheimer's disease.
I did not expect my lab will receive this amazing award again. Gabe Erion
@gabeerion
and Joe Janizek
@joejanizek
are MD/CSE PhD students. Safiye Celik
@safiscelik
and Scott Lundberg
@scottlundberg
were the winners in 2018 and 2017, respectively. I'm tremendously proud of you all!
The CoAI team from
@suinleelab
&
@UWMedicine
won the Madrona Prize recognizing excellence in research and commercial potential. CoAI is a machine learning method for predicting clinical outcomes to improve patient care. (2/6)
📢 Exciting news! The
@uw
Nathan Shock Center Basic Biology of Aging is hosting the “AI meets Biology of Aging” (AIMBA) virtual workshop on 5/22. Dive into cutting-edge research at the crossroads of AI and geroscience, with our stellar lineup of speakers!
Excited to share insights on the importance of explainable AI in drug discovery and biomedicine! Check out this
@medscape
story featuring my perspectives and my lab's latest research findings.
#ExplainableAI
#DrugDiscovery
Nicasia Beebe-Wang, my incredible final year Ph.D. student, will present her amazing work published in Nature Communications 2021 (in collaboration with a UW Allen School Prof. Sara Mostafavi
@sara_mostafavi
) at RECOMB 2022 soon!!
Here is another one accepted to NeurIPS 2020:
Ian Covert, Scott Lundberg, and Su-In Lee. "Understanding Global Feature Contributions through Additive Importance Measures."
Congratulations Ian Covert
@ianccovert
and Scott Lundberg
@scottlundberg
!
I'm incredibly proud to have worked with
@UW_MSTP
student
@joejanizek
, who has received his Computer Science
@uwcse
PhD from my lab! Wishing him all the best in his career. 🎓👏
📣 2024 Residency match results are out! 📣
We are so excited for each and every one of you! No matter where your journey takes you, we are so happy to have been part of this chapter.
💛🐾💜
Friends, please help spread the word by reposting!
Excited to announce the inaugural AIMBA (AI Meets Biology of Aging) 2024 meeting on May 22, *2024, 9-2pm PT, organized by
@UW
@NathanShockCtrs
!
Four keynote speakers:
@mariabrbic
,
@BrunetLab
, Vadim Gladyshev,
@james_y_zou
.
Excited to announce the inaugural AIMBA (AI Meets Biology of Aging) 2024 meeting on May 22, 2023, 9-2pm PT, organized by
@UW
@NathanShockCtrs
!
Featuring four keynote speakers: Profs
@mariabrbic
, Anne Brunet, Vadim Gladyshev,
@james_y_zou
.
Check it out!
@uw
's first MSTP student who completed Ph.D. in CS
@uwcse
, Gabe Erion
@gabeerion
, talks about our CoAI paper (Nature BME 2022).
UW AI helps ambulances and ICUs weigh emergency risks and cost considerations
Machine Learning in Computational Biology deadline extended until Wednesday 10/6, submit your 2 page abstract or 8 page paper at .
#MLCB2021
Please RT!
I know
@joejanizek
and Alex DeGrave, my 2nd/3rd
@uw
MSTP (MD/
@uwcse
PhD) students, all the hard work behind... Our Nature MI paper got cited >140 times since and helped the field go in the right direction!! I am tremendously proud of and feel *extremely* lucky to work with you!
Excited that this paper is published now! The biggest change since the pre-print is our finding confirming that models built on more carefully constructed datasets will generalize better to external hospitals
Our dynamic feature selection approach can further improve the interpretability of ML models. Super exciting to work with two exceptional co-first authors:
@uwcse
Ph.D. students,
@soham_gadgil
, and
@ianccovert
.
Excited to share our latest research featured in Allen School News! 🎉 Check out
@uwcse
Kristin Osborne's beautiful write-up on our recent
@natBME
paper with lead author Alex DeGrave and co-senior author
@RoxanaDaneshjou
.
#AI
image classifiers can help detect melanoma, but how they work has mostly been under wraps. A team led by
#UWAllen
’s
@suinleelab
&
@Stanford
clinicians devised a way to make their predictions medically understandable—and reveal where they miss the mark.
Working with the exceptional
@uwcse
@soham_gadgil
is a remarkable opportunity! Accepted to
#ICLR24
!
Check out
@soham_gadgil
's innovative idea on adaptively selecting features to enhance ML interpretation: … - a promising approach in medical diagnosis.
Always an honor to collaborate with brilliant students and colleagues! Check out this fantastic blog post about our publication in Nature Medicine, first authored by
@uwcse
student
@ChanwooKim_
and co-senior author
@RoxanaDaneshjou
:
Our new preprint - Erion, Janizek, and Sturmfels et al. by three alphabetically ordered first authors and Lundberg and Lee - on "Learning Explainable Models using Expected Gradients" to arXiv: .
Please visit our new MLCB website . MLCB registration is free and does not require paper submission. Due to space limit, we encourage to fill out the registration form.
Thank you so much everyone who pointed to me that something that I posted yesterday on this website did not come across as intended. I meant to say that our own level of effort is the only thing entirely within our control! (1/2)
The work of my MSTP (MD/CS Ph.D.) students, Alex DeGrave and Joe Janizek (alphabetically ordered co-first), "auditing" AI-based radiographic detection of COVID, got accepted to Nature Machine Intelligence.
I feel tremendously fortunate to be working with
@jmuiuc
on our Single-Cell Data Insights grants by
@ChanZuckerberg
. We will address fundamental problems in regulatory genomics using
#xai
techniques. Thank you
@CMUCompBio
for announcing our collaboration!
#UWAllen
@UW_MSTP
researchers led by
@suinleelab
used explainable
#AI
to assess how models predict COVID-19 status from chest x-rays. They discovered a tendency to focus on diagnostically irrelevant features—a shortcut that could lead to a misdiagnosis:
Thrilled to have the opportunity to work on this CZI
@ChanZuckerberg
grant with my very close friend
@suinleelab
. Our friendship has grown stronger over 15+ years, and Su-In and I can't wait to start working on this exciting and timely topic that is so dear to our heart! 1/2
Like many other people, I myself have had a significant hardship at important points in my life and career. I did not mean to imply that hard work is the only thing that affects success, academic, or otherwise. (2/2)
Check out our new pre-print on interpretable machine learning models for understanding drug synergy in AML! Wonderful collaboration with
@suinleelab
, tweetorial by star MD/PhD student
@joejanizek
below
👇
Seeing unfortunate terminations of friendships during this stressful pandemic, I thought about simple rules:
(1) Do not assume that you fully understand another person and their situation.
(2) Respect personal boundaries of other people.
(3) Do not be nosy. Focus on yourself.
Working with
@naxerova
@HMSGenetics
was an incredibly fascinating and inspiring experience! Check out what we learned by examining expression data through the lens of DNNs and interrogating them.
What happens if you study 50K cancer transcriptomes in a lot of analytic depth, and through the lens of a really good deep learning model?
I loved working on this fascinating project with
@suinleelab
! We dug into the data like maniacs😱
Read more 👇
It was a great honor to be the first speaker for the Pitt-CMU seminar series on ML in Medicine
@DBMI_Pitt
. Thank you so much for the great questions and discussions.
1/2 Very excited to launch our
#MLxMed
webinar series with a talk by amazing
@suinleelab
. The focus of the webinar series is on the application of ML in the Healthcare and will be held on every other Wed 3-4 PM via Zoom open to Pitt, CMU, UPMC.
#ML4H
#MLforHealth
@DBMI_Pitt
It was a mind-blowing experience to give a talk at the MLD seminar at the Carnegie Mellon University and meet with eminent scientists in the field of machine learning. Thank you,
@jmuiuc
Jian Ma and Barnabas Poczos for organizing my visit!
Finally, we highlight a paper from
@gabeerion
,
@joejanizek
,
@suinleelab
and colleagues (
@natBME
) on a cost-aware AI approach that optimizes the trade-off between clinical prediction performance and feature cost ().
If you want to know what your ViT pays attention to...you might not want to use attention values! Shapley values can do this better, and now they can even do it efficiently. Check out our new paper (ICLR spotlight) 🧵⬇️