Annotated
and Grouped Publication List – Murphy Group – January 30, 2017
Subjects
á Generative
Models of Subcellular Organization
á Active
Learning for Experimental Biology and Drug Development
á Subcellular
Pattern Unmixing
á Intelligent
Acquisition for Fluorescence Microscopy
á Subcellular Pattern
Analysis
o Tissues
o Cells
o Yeast
á Learning
Subcellular Sorting Pathways
á Content-based
Image Retrieval
Y. Li, T. D. Majarian, A. W. Naik, G. R. Johnson, and R. F. Murphy (2016) Point process
models for localization and interdependence of punctate cellular structures. Cytometry Part A 89:633-643
This paper builds on the Johnson et al study of punctate patterns by constructing point process models using different subcellular components as references. It shows that models built on cell and nuclear geometry and microtubule distribution are not improved by adding information on the spatial distribution of the endoplasmic reticulum.
K. T. Roybal, T. E. Buck, X. Ruan, B. H. Cho, D. J. Clark, R. Ambler, H. M. Tunbridge, J. Zhang, P. Verkade, C. WŸlfing, and R. F. Murphy (2016) Computational spatiotemporal analysis identifies WAVE2 and Cofilin as joint regulators of costimulation-mediated T cell actin dynamics. Science Signaling 9:rs3.
This close collaboration with Christoph WŸlfingÕs group involved constructing 4D Òspatiotemporal mapsÓ of the concentration of actin and eight of its regulators in T cells as they undergo immunological synapse formation. This was done by morphing each cell in each frame of thousands of movies into a standardized template.
R. M. Donovan, J.-J. Tapia, D. P. Sullivan, J.
R. Faeder , R. F. Murphy , M. Dittrich,
D. M. Zuckerman (2016) Unbiased Rare Event Sampling in Spatial
Stochastic Systems Biology Models Using A Weighted Ensemble Of Trajectories. PLoS Computational
Biology 12(2):e1004611.
This paper is the result of a collaboration
between investigators in the National Center for Multiscale
Modeling of Biological Systems. It describes using various cell
geometries, including those generated by CellOrganizer,
to efficiently carry out spatially accurate simulations of cell biochemistry.
G. R. Johnson, J. Li, A. Shariff, G.K.Rohde, and R.F. Murphy (2015) Automated Learning of Subcellular Pattern
Variation among Punctate Proteins and of a Generative Model of their
Distributions in Relation to Microtubules. PLoS Computational Biology 11(12): e1004614.
This paper uses images from the Human Protein Atlas to construct
generative models of the relationships between eleven punctate structures and
microtubules and shows that they can be used to accurately distinguish the
eleven patterns in three cell lines.
G. R. Johnson, T. E. Buck, D. P. Sullivan, G.
K. Rohde and R. F. Murphy (2015) Joint Modeling of Cell and Nuclear Shape Variation. Mol.
Biol. Cell, 26:4046-4056.
This paper provides the first statistical evidence for a relationship
between cell and nuclear shape, and shows that this relationship can be altered by gene alteration or drug addition. We also construct the first joint
generative model of the dynamics of cell and nuclear shape.
T.E. Buck,
J. Li, G.K. Rohde, and R.F. Murphy (2012) Towards the
virtual cell: Automated approaches to building models of subcellular
organization 'learned' from microscopy images. Bioessays 34:791-799.
J. Li, A. Shariff, M. Wiking, E. Lundberg, G.K. Rohde and R.F. Murphy (2012) Estimating microtubule distributions from 2D immunofluorescence microscopy images reveals differences among human cultured cell lines. PLoS ONE 7:e0050292.
This
paper builds generative models of microtubule patterns from 2D images for
different cultured cell lines using images from the Human Protein Atlas and
compares them.
R. F.
Murphy (2012) CellOrganizer: Image-derived Models of Subcellular Organization and Protein
Distribution. Methods in Cell Biology 110: 179-193.
R. F.
Murphy (2011) An active
role for machine learning in drug development. Nature
Chemical Biology 7:327-330.
T. Peng and R.F. Murphy (2011) Image-derived, Three-dimensional Generative Models of Cellular Organization. Cytometry Part A 79A:383-391.
This
paper describes extension of the initial 2D models of Zhao and Murphy (2007) to
3D.
A. Shariff,
R.F. Murphy, and G. Rohde (2011) Automated
Estimation of Microtubule Model Parameters from 3-D Live Cell Microscopy Images.
Proceedings of the 2011 IEEE
International Symposium on Biomedical Imaging (ISBI 2011), pp. 1330-1333.
This
paper describes modification of the microtubule model described below in order
to allow for estimation of free tubulin, and applies the model to images of
cells treated with and without nocodazole to
depolymerize microtubules. The
results are consistent with expectation.
R. F. Murphy (2010) Communicating
Subcellular Distributions. Cytometry Part A 77A:686-692.
This review provides a perspective on
methods for estimating pattern fractions and learning generative models. It addresses the critical problem of
representing information learned about subcellular organization for comparison
between cell and tissue types and for use in systems simulations.
A. Shariff,
G. K. Rohde and R. F. Murphy (2010) A Generative Model of Microtubule Distributions, and Indirect Estimation
of its Parameters from Fluorescence Microscopy Images. Cytometry 77A:457-466.
Methods have been described previously for learning models
of cell organization from microscope images in order to be able to synthesize
and combine subcellular distributions.
These methods involve direct estimation of the model parameters but for
some subcellular patterns (such as those of microtubules or microfilaments),
direct estimation is difficult due to large numbers of tangled fibers. We describe the first method for
indirectly learning a microtubule model and show that it produces results
consistent with current knowledge.
T. Peng, Wei Wang, G. K. Rohde, R. F. Murphy (2009) Instance-Based Generative Biological Shape Modeling. Proceedings of the 2009 IEEE International Symposium on Biomedical Imaging (ISBI 2009), pp. 690-693.
G. K. Rohde, W. Wang, T. Peng, and R.F. Murphy
(2008). Deformation-Based Nonlinear Dimension
Reduction: Applications To Nuclear Morphometry. Proceedings
of the 2008 IEEE International Symposium on Biomedical Imaging (ISBI 2008), pp. 500-503.
G. K. Rohde, A. Ribeiro, K. N. Dahl, and R. F. Murphy (2008). Deformation-based nuclear morphometry: capturing nuclear shape variation in HeLa Cells. Cytometry, 73A:341-350.
T. Zhao and R.F. Murphy (2007). Automated Learning of Generative Models for Subcellular Location: Building Blocks for Systems Biology. Cytometry 71A:978-990.
This was the first paper to describe the construction of generative
models of cell architecture directly from microscope images. It constructed models of cell and
nuclear shape and vesicular organelle size, shape and position.
A.W. Naik, J.D. Kangas,
D. P. Sullivan, and R. F. Murphy (2016) Active Machine Learning-driven Experimentation to Determine Compound
Effects on Protein Patterns,
eLife 5:e10047. doi:10.7554/eLife.10047.
This paper describes the first prospective
study to construct a predictive model of multiple drug
and target interactions using experiments selected solely under computer
control. The experiments were
carried out using liquid handling robotics and an automated microscope, and
performing only 28% of the experiments led to a model that was 92% accurate at
predicting the results of experiments (whether it had done them or not).
M. Temerinac-Ott,
A. W. Naik, and R. F. Murphy (2015) Deciding when to stop: Efficient experimentation to
learn to predict drug-target interactions. BMC
Bioinformatics 16:213 (also selected
for oral presentation in the Proceedings track of RECOMB 2015).
This paper uses four existing drug-target interaction datasets to show
that accurate models of these interactions can be constructed by active
learning without needing to do all experiments. It also provides the first evidence on
real datasets that the stopping rule
algorithm of the Naik et al paper below can be used
to estimate the accuracy of an actively learned model and decide when
experimentation can be stopped.
J.D. Kangas, A.W. Naik,
and R.F. Murphy (2014) Prediction of
Biological Responses Using Protein and Compound Features and
their Discovery using Active Learning. BMC
Bioinformatics 15:143. doi:10.1186/1471-2105-15-143
This paper describes the design of efficient regression models to
predict protein target responses to chemical compounds and shows that active
machine learning permits accurate models to be learned without doing all
experiments. We use a subset of PubChem data to test our combined approach, and use existing
features to describe the similarity among compounds and among protein
targets. The results show that 60%
of the ÒhitsÓ in the PubChem data could be discovered
while ÒdoingÓ only 3% of the possible experiments. This approach is complementary to that
in the Naik et al. paper below, which handles the
case where features are not available or reliable.
A. W. Naik,
J. D. Kangas, C. J. Langmead and
R. F. Murphy (2013) Efficient Modeling and Active Learning
Discovery of Biological Responses.
PLoS ONE 8: e83996. doi:10.1371/journal.pone.0083996
This paper characterizes new algorithms for active learning for drug
discovery in the absence of compound or target features. The algorithms seek to
learn the effects of many compounds on many targets, and address the case in
which the effect of a given compound on a given target is represented as one of
a number of different categorical phenotypes (rather than just as a score
measuring extent of an expected effect). We introduces measures of uniqueness
and responsiveness to characterize the nature of a given experimental
space, and show in simulated experiments that our active learner shows
significant improvement over using random choice and does so for essentially
all values of the uniqueness and responsiveness. We also introduce a stopping
rule approach for estimating the lower limit of the true accuracy of an
actively learned model, permitting decisions to be made about when to stop a campaign
of active learning-driven experimentation. Lastly, we show using Connectivity Map data that
accurate models of the effects of drugs on gene expression in various cell
lines can be constructed without the need to perform experiments for all
possible combinations of drugs and cell lines.
R. F.
Murphy (2011) An active
role for machine learning in drug development. Nature
Chemical Biology 7:327-330.
This commentary provides a perspective
on two critical areas in which machine learning methods are projected to
contribute to drug development and broader experimental biology: building image-drived models of subcellular organization, and using active
learning to avoid exhaustive experimentation of large experimental
spaces. It contains the first
proposal of active learning as a solution to the problem of considering all
possible interactions of many drugs and many targets.
R. F. Murphy (2010) Communicating
Subcellular Distributions. Cytometry Part A 77A:686-692.
This review provides a perspective on
methods for estimating pattern fractions and learning generative models. It addresses the critical problem of
representing information learned about subcellular organization for comparison
between cell and tissue types and for use in systems simulations.
L. P. Coelho, T. Peng, and R. F.
Murphy (2010) Quantifying the
distribution of probes between subcellular locations using unsupervised pattern
unmixing. Bioinformatics 26:i7-i12
(Proceedings of 18th Annual International Conference on
Intelligent Systems in Molecular Biology; only 19% of submitted papers
accepted).
Supervised
approaches to pattern unmixing require examples of
images for proteins that are found in only one fundamental subcellular pattern
(e.g., organelle). When analyzing
protein images on a proteome scale, the patterns may not all be known and/or
proteins that are only present in each of these patterns may not be available. This paper described the first system
for unsupervised unmixing
of patterns, that is, simultaneously finding the underlying patterns and
estimating the fraction of each protein in each.
T. Peng,
G.M.C. Bonamy, E. Glory-Afshar,
D. R. Rines, S. K. Chanda,
and R. F. Murphy (2010) Determining the distribution of probes between different
subcellular locations through automated unmixing of
subcellular patterns. Proc. Natl.
Acad. Sci. U.S.A. 107:2944-2949.
Proteins may be found in more than
one subcellular location, but previous automated systems to classify images by
their patterns could not estimate the amount in each. This paper is the first demonstration of
the ability to unmix subcellular patterns in
microscope images. It was chosen
for a Highlights Track presentation at ISMB 2010.
T. Zhao, M. Velliste, M.V. Boland,
and R.F. Murphy (2005). Object Type
Recognition for Automated Analysis of Protein Subcellular Location. IEEE Trans. Image Proc. 14:1351-1359
C. Jackson,
E. Glory, R. F. Murphy and J. Kovacevic (2011) Model building and intelligent acquisition with application to protein
subcellular location classification. Bioinformatics
27:1854-1859.
This paper describes a model of object
dynamics and an algorithm for acquiring images of a given sample to efficiently
learn the model parameters.
C. Jackson,
R. F. Murphy, and J. Kovacevic (2009) Intelligent
Acquisition and Learning of Fluorescence Microscope Data Models. IEEE
Trans Image Proc. 18:2071-2084.
C. Jackson, R.F. Murphy
and J. Kovacevic (2007). Efficient
Acquisition and Learning of Fluorescence Microscopy Data Models. Proceedings
of 2007 IEEE International Conference on Image Processing, pp. VI-245-VI-248.
A. Kumar, A. Rao, S.
Bhavani, J.Y. Newberg, R. F. Murphy (2014) Automated
Analysis of Immunohistochemical Images Identifies
Candidate Location Biomarkers for Cancers. Proc. Natl. Acad. Sci. U.S.A.
111:18249-18254.
This paper describes the construction
of a system for measuring changes in subcellular location between normal and
cancerous tissues (the subject of U.S.
patent number 9,092,850) and uses it to
identify proteins that are Òlocation biomarkersÓ of different types of cancers.
A. Rao and
R.F. Murphy (2011) Determination of Protein Location Diversity Via Analysis of Immunohistochemical Images from the Human Protein Atlas. Proceedings of the 2011 IEEE International
Symposium on Biomedical Imaging
(ISBI 2011), 1727-1729.
E. Glory-Afshar, E. Garcia Osuna, B. Granger, and R. F. Murphy (2010) A Graphical Model To Determine The Subcellular Protein Location In Artificial Tissues. Proceedings of the 2010 IEEE International Symposium on Biomedical Imaging (ISBI 2010), pp. 1037-1040.
E. Glory, J.
Newberg, and R.F. Murphy (2008). Automated
Comparison Of Protein Subcellular Location Patterns Between Images Of Normal
And Cancerous Tissues. Proceedings of
the 2008 IEEE International Symposium on Biomedical Imaging (ISBI 2008), pp. 304-307.
J. Newberg and R.F. Murphy (2008). A Framework for the Automated Analysis of Subcellular Patterns in Human Protein Atlas Images. J. Proteome Res. 7: 2300-2308.
L. P. Coelho, J. D. Kangas,
A. Naik, E. Osuna-Highley,
E. Glory-Afshar, M. Fuhrman, R. Simha,
P. B. Berget, J. W. Jarvik,
and R. F. Murphy (2013) Local Features Provide Better Generalization of Subcellular Location Classifiers to New Proteins. Bioinformatics 29: 2343-2349.
This paper provides a new perspective on the problem
of classifying subcellular patterns.
Previous work, beginning with our initial framing of the problem in Boland et al
(1997), Boland et al
(1998) and Boland
& Murphy (2001), considered the problem of recognizing organelle
patterns by measuring recognition of new images of the same marker proteins
that had been used for training. However,
this does not provide a good estimate of performance for classifying images of
new proteins localized to the same organelle, which may not have exactly the
same pattern as those used for training.
Since previous image collections to not allow assessment of this
performance, we describe new open access image collections containing multiple
proteins that localize to each major.
We also describe modifications to local feature methods that incorporate
information from reference channels and show that these new features, combined
with previously described features, provide improved performance on this
task.
J. Li, J.Y. Newberg, M. UhlŽn, E. Lundberg, and R.F. Murphy (2012) Automated
Analysis and Reannotation of Subcellular Locations in
Confocal Images from the Human Protein Atlas. PLoS ONE 7:e0050514.
J. Li, L. Xiong, J. Schneider, and R.F. Murphy (2012) Protein Subcellular Location Pattern Classification in Cellular Images Using Latent Discriminative Models. Bioinformatics 28, i32-39
Y. Hu, E. Garcia Osuna, J. Hua, T. S. Nowicki,
R. Stolz,
C. McKayle
and R. F. Murphy (2010) Automated Analysis of Protein Subcellular
Locations in Time Series Images.
Bioinformatics 26:1630-1636.
Most work on automatically classifying
subcellular patterns uses static images and is unable to distinguish proteins
by their dynamic behavior. This
paper describes a number of approaches for calculating features to describe
variation in location over time, and shows that these features allow better
discrimination between protein patterns.
J. Y. Newberg, J. Li, A. Rao, E. Lundberg, F. Ponten, M. Uhlen and R. F. Murphy (2009) Automated
Analysis Of Human Protein Atlas Immunofluorescence Images. Proceedings of the 2009 IEEE International
Symposium on Biomedical Imaging
(ISBI 2009), pp. 1023-1026.
S. Huh, D. Lee and R. F. Murphy (2009) Efficient framework for automated classification of subcellular patterns in budding yeast. Cytometry 75A:934-940.
S.-C. Chen, T. Zhao,
G. J. Gordon, and R. F. Murphy (2007). Automated Image
Analysis of Protein Localization in Budding Yeast. Bioinformatics
23:i66-i71
T. Lin, Z. Bar-Joseph, and R. F. Murphy (2011) Learning Cellular
Sorting Pathways Using Protein Interactions and Sequence
Motifs. Journal of Computational
Biology 18: 1709-1722.
T. Lin, Z. Bar-Joseph, and R. F. Murphy (2011) Learning Cellular Sorting Pathways Using
Protein Interactions and Sequence Motifs. Lecture Notes in Bioinformatics (Proceedings
of RECOMB 2011) 6577:204-221.
This paper (presented at RECOMB and
published in slightly edited form in an issue of the Journal of Computational Biology featuring selected papers) uses
both known motifs as well as motifs learned as described in Lin et al (2010)
(below) in combination with data on protein-protein interaction to learn a
sorting model for subcellular localization.
T. Lin, R.F. Murphy, and
Z. Bar-Joseph (2010) Discriminative
Motif Finding for Predicting Protein Subcellular Localization. IEEE/ACM Transactions on Computational Biology
and Bioinformatics 8:441-51.
Many systems for predicting
subcellular location of proteins using sequence motifs have been described, but
this paper describes the first approaches for learning these motifs given just
sequences and locations. The system can achieve results comparable to the best current
predictors but on the much harder task of learning motifs as well.
B.H. Cho,
I. Cao-Berg, J.A. Bakal, and R.F. Murphy (2012) OMERO.searcher: Content-based image search for microscope images. Nature Methods 9:633-634.
This paper showed
that Subcellular
Location Features described previously (see Subcellular Pattern Analysis) can be
used to find images in an OMERO database that match the pattern of a
user-supplied image of an unknown cell pattern. The open source software is now included
in the OMERO distribution.
R. F. Murphy (2016) Building Cell Models and
Simulations from Microscope Images. Methods 96:33-39
K.T. Roybal, P. Sinai, P. Verkade,
R. F. Murphy, and Christoph WŸlfing
(2013) The actin-driven
spatiotemporal organization of signaling in T cells activated by antigen
presenting cells. Immunological Reviews 256: 133-147.
T.E. Buck,
J. Li, G.K. Rohde, and R.F. Murphy (2012) Towards the
virtual cell: Automated approaches to building models of subcellular
organization 'learned' from microscopy images. Bioessays 34:791-799.
K.W. Eliceiri, M.R. Berthold, I.G. Golberg,
L. Ibanez, B.S. Manjunath, M.E. Martone,
R.F. Murphy, H. Peng, A.L. Plant, B. Roysam, N. Stuurmann, J.R.Swedlow, P. Tomancak, and
A.E. Carpenter (2012) Biological Imaging Software Tools. Nature Methods 9:697-710.
R. F.
Murphy (2012) CellOrganizer: Image-derived Models of Subcellular Organization and Protein
Distribution. Methods in Cell Biology 110: 179-193.
R. F.
Murphy (2011) An active
role for machine learning in drug development. Nature
Chemical Biology 7:327-330.
This commentary provides a perspective
on two critical areas in which machine learning methods are projected to
contribute to drug development and broader experimental biology: building image-derived
models of subcellular organization, and using active learning to avoid
exhaustive experimentation of large experimental spaces. Our work on these areas is described in
separate sections above.
R. F. Murphy (2010) Communicating
Subcellular Distributions. Cytometry Part A 77A:686-692.
This review provides a perspective on
methods for estimating pattern fractions and learning generative models. It addresses the critical problem of
representing information learned about subcellular organization for comparison
between cell and tissue types and for use in systems simulations.
A. Shariff, J. Kangas,
L.P. Coelho, S. Quinn and R.F. Murphy (2010) Automated Image
Analysis for High Content Screening and Analysis. J. Biomolec. Screening 15:726-734.
L. P. Coelho, E. Glory-Afshar, J. Kangas, S. Quinn, A. Shariff, and R. F. Murphy (2010) Principles of Bioimage Informatics: Focus on machine learning of cell patterns. Lecture Notes in Computer Science 6004:8-18.