My
research lies in the interface of machine learning, combinatorial and numerical
optimization and some data-rich areas such as computational
biology. I am interested in studying theoretically sound and empirically
efficient machine learning and optimization algorithms to analyze, interpret,
and integrate high volumes of multi-dimensional and heterogeneous data and
based upon which building effective predictive models. On one hand, I would
like to develop computational methods for addressing a specific challenging
problem by taking full advantage of all available data and domain knowledge. On
the other hand, inspired by real-world applications I would like to study novel
machine learning and optimization algorithms that is widely applicable.
1. Invited to give a keynote talk at ISMB 3DSIG 2019.
2. One 12-year old paper entitled “Pairwise Global Alignment of
Protein Interaction Networks by Matching Neighborhood Topology” received the Test-of-Time Award at RECOMB 2019.
See here
for news.
3. Our deep
learning method for protein contact prediction and tertiary structure
prediction is ranked 1st and 2nd (in server category), respectively, in
CASP13 (the 13th Critical Assessment of Protein Structure
Prediction). All the top predictors in contact prediction and template-free
modeling adopted the deep learning algorithm we first published in “Accurate De Novo Prediction of Protein Contact Map
by Ultra-Deep Learning Model”.
4. Our deep
learning paper for contact prediction entitled “Accurate De Novo Prediction of Protein Contact Map by Ultra-Deep
Learning Model” won the 2018 PLoS Computational Biology Research Prize in the
category of Breakthrough/Innovation.
5.
Invited
to give a keynote talk at IEEE ICCABS on
Oct 19, 2017.
6.
Our
deep learning method for protein contact prediction is ranked
1st in CASP12, the 12th Critical Assessment of Protein
Structure Prediction. See here for a
ranking list and our
paper for technical details. This work is also covered by The
Economist.
7. Science reports our
protein structure prediction server RaptorX. See the
report for more details.
8. Our paper
entitled "CNVnet: Combining Sparse Learning and
Biological Networks to Capture Joint Effect of Copy Number Variants" has
received the ACM SIGBio Best Student Paper Award
during the ACM-BCB
2014. The authors
are Mr. Zhiyong Wang (PhD student), Prof. Xinghua Shi
(UNCC) and Jinbo Xu.
9. The first three
PhD students in my group have landed their desirable jobs: Dr. Jian Peng
(tenure-track faculty at UIUC CS), Dr. Feng Zhao (finance industry in New
York), and Zhiyong Wang (Google).
10. Our presentation
entitled MRFalign: protein remote homology detection through
alignment of Markov Random Fields won the
Warren DeLano Award for Structural Bioinformatics and Computational Biophysics in
ISMB 3DSIG 2014. ISMB is one of the top 2 bioinformatics conferences.
11. Our paper
entitled MRFalign: protein remote homology detection through
alignment of Markov Random Fields won the best
paper award in RECOMB 2014, one of the top 2 bioinformatics conferences.
12. Our PhD student
Mr. Jianzhu Ma won the best
poster award in the 2013 Zing conference for protein and RNA structure
analysis. The poster is about protein homology detection by aligning two Markov
Random Fields.
13. Our protein
structure prediction program RaptorX
was ranked among top 10 for the CASP10 hardest template-based modeling targets.
RaptorX is the only server among the top 10 groups. The other 9 are human
groups, which can make use of server results.
14. In ISMB 3DSIG 2012,
Dr. Sheng Wang won the Warren DeLano Award for Structural
Bioinformatics and Computational Biophysics for his work entitled protein structure alignment beyond spatial
proximity, which is implemented as DeepAlign
(available at http://ttic.uchicago.edu/~jinbo/software.htm).
15. Our protein
structure prediction program RaptorX was ranked No.1 among
~80 CASP9 participating servers for the 50 hardest TBM (template-based
modeling) targets in CASP9 and voted as one of the most interesting and
innovative methods by the CASP9 community. Our team was invited to give three
talks at the CASP9 meeting.
16. Our PhD student
Mr. Jian Peng wins the prestigious Microsoft Research
PhD Fellowship 2010. Only 10 out of 176 applications received this award
in 2010.
17. Our poster
entitled Boosting Protein Threading
Accuracy won the Best Poster Award in RECOMB
2009, one of the best conferences in the field of Computational Biology and
Bioinformatics.
18. The following
poster won the Best Poster Award in CASP8
conference held in
J. DeBartolo, G.
Hockey, F. Zhao, J. Peng, A. Augustyn, A. Adhikar, J. Xu, K. F. Freed and T. R. Sosnick.
Structure
prediction combining the template-based RAPTOR algorithm with the ItFix ab initio method.
16. Our protein structure prediction program RAPTOR ranked No.2-No.4 among all automated programs in CASP8, according to several assessments.
1.
Jinbo
Xu. NSF/CCF AF-1618648. AF:III: small: Convex optimization for protein-protein interaction network
alignment. Total cost $300k, 7/1/2016-6/30/2019.
2.
Jinbo
Xu. NSF/BIO-1564955. ABI Development:
Developing RaptorX Web Portal for Protein Structure and Functional Study. Total cost $557k, 7/15/2016-6/30/2019.
3.
Jinbo
Xu. NIH R01GM089753. New computational methods for data-driven
protein structure prediction. Total cost ~$2.65m, 5/1/2010-8/31/2019
4.
Jinbo
Xu. NSF/CCF AF-1149811 (CAREER award). Exact
and approximate algorithms for 3D structure modeling of protein-protein
interactions. Total cost $500k, 7/1/2012-6/30/2017
5.
Jinbo
Xu. NSF/BIO DBI-1262603. Continued development of RaptorX for protein
structure and function prediction. Total cost ~$540k, 7/1/2013-6/30/2016
6.
Jinbo
Xu. NSF/BIO DBI-0960390: Algorithm and
web server for low-homology protein threading. Total cost ~$408k,
7/1/2010-6/30/2013
7.
Tobin
Sosnick (PI), Karl Freed, and Jinbo Xu. NIH
R01GM081642. Protein structure refinement
using novel move set. My share ~$200k, 08/2007-08/2010
8.
Bonnie Berger (PI),
Jinbo Xu and Jadwiga Bienkowska. NIH R01GM081871A1. Prediction of
protein interactome. My share ~$100k, 12/2007-12/2012
9.
Bonnie Berger (PI),
Jinbo Xu and Jadwiga Bienkowska. NIH R01GM081871A1. Prediction of
protein interactome. My share ~$100k, 12/2007-12/2012