June 1, 2016 (Vol. 36, No. 11)

Richard A. A. Stein M.D., Ph.D.

The Interconnected Systems behind Disease Processes Are Being Unraveled By Systems Biology

Biotechnology, molecular biology, and genetic engineering share a firm foundation—a massive body of knowledge about individual cellular and subcellular components. At the same time, these disciplines are limited in the same way. They can rise only so high before they strain to accommodate whole systems and their emergent behaviors.

To help the life sciences reach greater heights, biologists have been exploring interdisciplinary efforts. Traditional life science disciplines are being buttressed by engineering and computational and mathematical sciences. The result: a rising edifice called systems biology. It reflects how specific biological problems may be interconnected.

“We cannot reduce disease to one or two genes. Instead, we need to look at how pathways work together and interact in cells,” says Olivier Elemento, Ph.D., associate professor of physiology and biophysics at Weill Cornell Medical College. “Only by embracing this complexity can we understand disease.”

Dr. Elemento heads a laboratory that focuses on identifying the cellular targets of small molecules, a task of critical importance for molecular biology, pharmacology, and drug design. However, identifying the targets of small molecules is complicated by several factors, including the ability of some compounds to target multiple proteins.

“Our strategy,” informs Dr. Elemento, “relies on using a combination of computational biology and systems biology.”


At Weill Cornell Medical College, the laboratory of Olivier Elemento, Ph.D., combines computational biology and systems biology to identify mechanisms of drug resistance. The laboratory uses ultrafast genome and DNA sequencing, proteomics, high-performance computing, mathematical modeling, and machine learning to characterize regulatory networks.

Overcome Drug Resistance

Dr. Elemento’s team used a computational/systems biology strategy when it demonstrated that transcriptome sequencing could be used to identify mechanisms of drug action and resistance. The team began by establishing a cell line that is sensitive to a small compound. Next, the team isolated resistant clones. Finally, the team compared the transcriptomes of the resistant clones to the transcriptome of the parental cell population to identify mutations that could account for the resistance.

According to Dr. Elemento’s team, mutations common to more than one clone were useful in revealing a drug’s physiological target and indirect resistance mechanisms. Following such mutational clues is likely to be fruitful even if prior knowledge about the identity of a drug target or its cellular location is lacking. One of the most common mechanisms of acquired drug resistance is modification of the drug target to prevent binding.

In another study, one in which a library of over 10,000 different compounds was screened in a drug-sensitive fission yeast model, Dr. Elemento’s team identified Cutin-1, an inhibitor of mitosis and chromosome segregation. Moreover, the team revealed that this protein targets the dehydratase subunit of the fatty acid synthase.

“Advances in technology have made it easy for us to perform genome-wide sequencing and examine transcriptomes, mutations, or the binding of small molecules to the genome,” notes Dr. Elemento. “We have been able to delve into various layers of gene expression and biochemical communication. But there are still layers that contain under-explored depths. One of these layers is post-translational modification.”

Advances in molecular and imaging approaches have helped dissect the individual components of signaling pathways and visualize their structural and functional details, and they have been instrumental for characterizing networks and for advancing system biology. However, one of the obstacles in modeling networks lies in the uncertainty about biochemical parameters.

“We are reaching the point where we have a few frameworks that are working well,” asserts Dr. Elemento. “And we can do even better. There is a lot of hope in using Bayesian approaches to model signaling networks.”

Recognize Heterogeneity

 “Systems biology is not just about measuring as many things as possible,” says lya Shmulevich, Ph.D., a professor at the Institute for Systems Biology. “It is also about using mathematical and computational modeling to make predictions about the system.”

Historically, most systems biology efforts have focused on generating models at the cellular level to provide a better understanding of intracellular processes. For example, models of genetic regulatory networks helped understand how intracellular genes and proteins interact with each other to govern cellular behaviors and decisions.

Yet, as Dr. Shmulevich emphasizes, it is not enough to model only what happens inside individual cells. Pathological processes involve the interplay of large numbers of cells, which collectively shape the phenotype, but the intracellular networks of the individual cells in these heterogeneous populations exhibit substantial cell-to-cell differences. Another consideration is that the stromal environment of tumors contains multiple distinct cell types, including normal cells such as infiltrating immune cells, endothelial cells, and adipose cells.

“These cells interact with one another physically and chemically,” explains Dr. Shmulevich. Interactions between cells may occur over short or long distances, and they modulate the disease process and the response to therapy. “It is not enough to model genetic networks inside cancer cells,” Dr. Shmulevich continues. “We need to acknowledge the complexity of the tumor as an ecosystem of many cells and different cell types that are interacting on different scales, and begin to model it.”

A key effort in Dr. Shmulevich’s laboratory focuses on The Cancer Genome Atlas (TCGA), a project that collects and integrates vast datasets from multiple cancer types. “So far, TCGA has collected data from 33 different primary tumor types from various organs,” notes Dr. Shmulevich. Each sample that was donated to the project generated a wealth of information—DNA, RNA, epigenetic, copy number variation, microRNA, protein, and clinical data.

“These data,” declares Dr. Shmulevich, “helped create a comprehensive measurement for all these cancer types, allowing comparisons that help understand molecular processes that are involved in cancer onset and progression.”

Dr. Shmulevich’s lab is actively involved in interrogating the cancer immune response from a systems biology perspective. Various experimental approaches have identified and characterized the immune cells that are present in malignant tumors, and the global systems analysis promises to translate the basic research into actionable clinical interventions.

“From the available data,” informs Dr. Shmulevich, “we can computationally work out what the immune response is in the malignancy, compare different cancer types and subtypes, and learn about which immune modulators or which particular mechanisms could be used therapeutically.”

Immunotherapies have shown promising results in cancer therapy. For example, PDL-1 checkpoint inhibitors recently emerged as some of the most successful therapeutic approaches in some malignancies.

“However, these types of therapy work only in a small percentage of the patients,” cautions Dr. Shmulevich. “It is important to understand how to predict patients that will benefit and how to choose the best therapy for each patient in a personalized way.”

Capture Adaptive Dynamics

In textbooks, metabolism is usually presented as a static process, one that is shaped by the availability of specific nutrients at particular moments. This view, the so-called canonical view, is too static, complains Markus Ralser, Ph.D., group leader in the department of biochemistry at the University of Cambridge and the Francis Crick Institute.

“The static view is not compatible with the need for cells to adapt to their ever-changing environments if they are to maintain their chemical make-up,” explains Dr. Ralser. “We are looking at metabolism in context of how cells adapt to stress conditions and to changes in their immediate environment.” These environmental changes include fluctuations in temperature, which have a huge impact on the speed of biochemical reactions, exposures to toxins and oxidants, changes in oxygen availability, and nutritional factors.

Dr. Ralser’s laboratory focuses on the flexibility with which cells adapt. In particular, the laboratory is working to understand how this flexibility depends on carbon and amino acid metabolism.

“We started to look at how metabolic differences impact gene expression,” recalls Dr. Ralser. “We found that metabolism is very tightly bound to how the genome and proteome responds to any kind of perturbation, but there is also a lot of metabolic self-regulation.”

In recent studies, Dr. Ralser and colleagues revealed that environmental changes can activate signaling pathways and elicit specific genomic responses, such as copy number changes, as a way to adapt to environmental conditions. Signaling- and genome-level adaptations to environmental conditions are difficult to study, says Dr. Ralser, because they proceed with very different kinetics.

“While metabolite concentrations change in seconds, and while transcripts and proteins change in minutes, genomic changes take hours, explains Dr. Ralser. “It is very difficult to predict which part of the picture is missed at a given moment.”

Predict Adverse Drug Effects

Scientists at the Icahn School of Medicine at Mount Sinai hope to accelerate the discovery of new therapies and diagnostics through Big Data approaches. For example, to study how human cells respond to drugs, these scientists are mining gene-expression data from the Library of Integrated Network-based Cellular Signatures (LINCS), which catalogues information about cellular responses to various genetic and environmental stressors, including more than 20,000 small molecule compounds.

In one project, LINCS information was examined in light of the chemical structural features of selected drugs. “By linking human phenotypes with molecular data, we have shown that we can improve the prediction of adverse drug effects,” says Avi Ma’ayan, Ph.D., professor of pharmacology and systems therapeutics at Mount Sinai.

“This project has generated large transcriptomics, proteomics, and imaging datasets from human cell lines perturbed with different small molecules and other types of factors, sometimes in combination,” explains Dr. Ma’ayan. “This project proposes to shed light on how human cells respond to perturbations and to help understand differences and commonalities in the response between the cell types.”

An advantage of this strategy is that it requires only information about the chemical structure and the gene-expression profile, but not necessarily information about the intracellular targets of a molecule. Nevertheless, this approach does not capture adverse effects that result from actions at the level of organs or systems that do not involve cellular receptors, and it is challenging to use it to interrogate drugs that are first metabolized in the body.

Dr. Ma’ayan and colleagues have made their application available as a searchable web portal. By using this portal, investigators can interrogate compounds of interest for which relevant data is available.


The Ma’ayan Laboratory of Computational Systems Biology at Mount Sinai studies intracellular regulatory networks to characterize processes such as differentiation, de-differentiation, apoptosis, and proliferation. To do so, it applies graph-theory algorithms, machine-learning techniques, and dynamical modeling. In one project, the Ma’ayan Laboratory used the LINCS database to predict that kenpaullone might serve as a therapeutic against Ebola.

Capture Strain-to-Strain Variation

“Our technology involves understanding the antibody response upon exposure to an infectious agent,” says Philip L. Felgner, Ph.D., professor of infectious disease and immunology at the University of California, Irvine. This technology may help scientists deal with strain-to-strain variation among microorganisms.

A major challenge in understanding infectious diseases such as malaria concerns the sequence variations that occur naturally in the environment, and the role of strain variants in immune recognition and susceptibility to infection. This challenge may be met, however, through the application of advanced sequencing technologies, which have made it possible to generate whole genomes within hours or days. A marked variability in the genetic makeup was found even for the same organism, when isolated from different infections from the same species and different regions around the world.

“In the environment, microorganisms are much more different than in the lab,” notes Dr. Felgner. “Our strategy can help us understand how variable these microorganisms are.” Strain-to-strain variation is reflected in the antibody repertoires that are elicited by infections with individual microbial strains, and it has diagnostic and therapeutic implications.

To capture information about antibody profiles elicited as a result of infection, Dr. Felgner and colleagues developed a protein microarray platform on which all the individual proteins from a microorganism of interest can be printed on a microarray chip. “Our platform is powerful because it can generate all the individual proteins from an infectious microorganism and print them on a chip,” explains Dr. Felgner. “We can use the chip to identify the antibodies that people make when becoming infected.”

As part of this project, Dr. Felgner and colleagues generated about 60,000 proteins from 35 different microorganisms. “Using a high-throughput method, one researcher can clone 384 genes a day using this approach, allowing all the protein-encoding genes from a bacterium such as Mycobacterium tuberculosis to be cloned in a few weeks,” asserts Dr. Felgner. A high-throughput expression system subsequently performs in vitro transcription and translation, and it allows the protein array to be printed on 384-well plates.

“We generated an inventory of protein arrays that allows us to go into the field and probe specimens from people with natural exposure and clinical syndromes,” states Dr. Felgner. “We can also use the approach in vaccine clinical trials to better understand the antibody response that occurs as a result of exposure.”

A major effort in Dr. Felgner’s laboratory focuses on understanding the antibody response in Q fever. While a Q fever vaccine exists and confers protection, its use has been accompanied by a high rate of allergic reactions. “Once we learn more about the antibody response, we can develop a vaccine that does not use whole killed organisms but only one or a few antigens,” says Dr. Felgner. “Such a vaccine might show a superior efficacy and safety profile.” 


At the University of California, Irvine, in the laboratory of Philip L. Felgner, Ph.D., microarray technology is being used to assess strain-to-strain variation among microorganisms. Shown here are two microarrays with 4,500 proteins from the malaria parasite. The image on the left shows a U.S. malaria-naïve adult, while the image on the right shows a naturally exposed resident of Mali West Africa. Both samples used a probe with sera followed by a fluorescent secondary antibody.

Previous articleStemCells Winding Down Operations after Ending Phase II Study
Next articleMulti-Dimensional Impact of the Public-Private Center for Translational Molecular Medicine