Personalized disease signatures through information-theoretic compaction of big cancer data.

Every individual cancer develops and grows in its own specific way, giving rise to a recognized need for the development of personalized cancer diagnostics. This suggested that the identification of patient-specific oncogene markers would be an effective diagnostics approach. However, tumors that are classified as similar according to the expression levels of certain oncogenes can eventually demonstrate divergent responses to treatment. This implies that the information gained from the identification of tumor-specific biomarkers is still not sufficient. We present a method to quantitatively transform heterogeneous big cancer data to patient-specific transcription networks. These networks characterize the unbalanced molecular processes that deviate the tissue from the normal state. We study a number of datasets spanning five different cancer types, aiming to capture the extensive interpatient heterogeneity that exists within a specific cancer type as well as between cancers of different origins. We show that a relatively small number of altered molecular processes suffices to accurately characterize over 500 tumors, showing extreme compaction of the data. Every patient is characterized by a small specific subset of unbalanced processes. We validate the result by verifying that the processes identified characterize other cancer patients as well. We show that different patients may display similar oncogene expression levels, albeit carrying biologically distinct tumors that harbor different sets of unbalanced molecular processes. Thus, tumors may be inaccurately classified and addressed as similar. These findings highlight the need to expand the notion of tumor-specific oncogenic biomarkers to patient-specific, comprehensive transcriptional networks for improved patient-tailored diagnostics.

Statistical thermodynamics of transcription profiles in normal development and tumorigeneses in cohorts of patients.

Experimental biology is providing the distribution of numerous different biological molecules inside cells and in body fluids of patients. Statistical methods of analysis have very successfully examined these rather large databases. We seek to use a thermodynamic analysis to provide a physical understanding and quantitative characterization of human cancers and other pathologies within a molecule-centered approach. The key technical development is the introduction of a Lagrangian. By imposing constraints the minimal value of the Lagrangian defines a thermodynamically stable state of the cellular system. The minimization also allows using experimental data measured at a number of different conditions to evaluate the steady-state distribution of biomolecules such as messenger RNAs. Thereby the number of effectively accessible quantum states of biomolecules is determined from the experimentally measured expression levels. With the increased resolution provided by the minimization of the Lagrangian one can differentiate between normal and diseased patients and further between disease subtypes. Each such refinement corresponds to imposing an additional constraint of biological origin. The constraints are the unbalanced ongoing biological processes in the system. MicroRNA expression level data for control and diseased lung cancer patients are analyzed as an example.

Computational surprisal analysis speeds-up genomic characterization of cancer processes.

Surprisal analysis is increasingly being applied for the examination of transcription levels in cellular processes, towards revealing inner network structures and predicting response. But to achieve its full potential, surprisal analysis should be integrated into a wider range computational tool. The purposes of this paper are to combine surprisal analysis with other important computation procedures, such as easy manipulation of the analysis results--e.g. to choose desirable result sub-sets for further inspection--, retrieval and comparison with relevant datasets from public databases, and flexible graphical displays for heuristic thinking. The whole set of computation procedures integrated into a single practical tool is what we call Computational Surprisal Analysis. This combined kind of analysis should facilitate significantly quantitative understanding of different cellular processes for researchers, including applications in proteomics and metabolomics. Beyond that, our vision is that Computational Surprisal Analysis has the potential to reach the status of a routine method of analysis for practitioners. The resolving power of Computational Surprisal Analysis is here demonstrated by its application to a variety of cellular cancer process transcription datasets, ours and from the literature. The results provide a compact biological picture of the thermodynamic significance of the leading gene expression phenotypes in every stage of the disease. For each transcript we characterize both its inherent steady state weight, its correlation with the other transcripts and its variation due to the disease. We present a dedicated website to facilitate the analysis for researchers and practitioners.

miRNA and mRNA cancer signatures determined by analysis of expression levels in large cohorts of patients.

Toward identifying a cancer-specific gene signature we applied surprisal analysis to the RNAs expression behavior for a large cohort of breast, lung, ovarian, and prostate carcinoma patients. We characterize the cancer phenotypic state as a shared response of a set of mRNA or microRNAs (miRNAs) in cancer patients versus noncancer controls. The resulting signature is robust with respect to individual patient variability and distinguishes with high fidelity between cancer and noncancer patients. The mRNAs and miRNAs that are implicated in the signature are correlated and are known to contribute to the regulation of cancer-signaling pathways. The miRNA and mRNA networks are common to the noncancer and cancer patients, but the disease modulates the strength of the connectivities. Furthermore, we experimentally assessed the cancer-specific signatures as possible therapeutic targets. Specifically we restructured a single dominant connectivity in the cancer-specific gene network in vitro. We find a deflection from the cancer phenotype, significantly reducing cancer cell proliferation and altering cancer cellular physiology. Our approach is grounded in thermodynamics augmented by information theory. The thermodynamic reasoning is demonstrated to ensure that the derived signature is bias-free and shows that the most significant redistribution of free energy occurs in programming a system between the noncancer and cancer states. This paper introduces a platform that can elucidate miRNA and mRNA behavior on a systems level and provides a comprehensive systematic view of both the energetics of the expression levels of RNAs and of their changes during tumorigenicity.

On a fundamental structure of gene networks in living cells.

Computers are organized into hardware and software. Using a theoretical approach to identify patterns in gene expression in a variety of species, organs, and cell types, we found that biological systems similarly are comprised of a relatively unchanging hardware-like gene pattern. Orthogonal patterns of software-like transcripts vary greatly, even among tumors of the same type from different individuals. Two distinguishable classes could be identified within the hardware-like component: those transcripts that are highly expressed and stable and an adaptable subset with lower expression that respond to external stimuli. Importantly, we demonstrate that this structure is conserved across organisms. Deletions of transcripts from the highly stable core are predicted to result in cell mortality. The approach provides a conceptual thermodynamic-like framework for the analysis of gene-expression levels and networks and their variations in diseased cells.

Protein signaling networks from single cell fluctuations and information theory profiling.

Protein signaling networks among cells play critical roles in a host of pathophysiological processes, from inflammation to tumorigenesis. We report on an approach that integrates microfluidic cell handling, in situ protein secretion profiling, and information theory to determine an extracellular protein-signaling network and the role of perturbations. We assayed 12 proteins secreted from human macrophages that were subjected to lipopolysaccharide challenge, which emulates the macrophage-based innate immune responses against Gram-negative bacteria. We characterize the fluctuations in protein secretion of single cells, and of small cell colonies (n = 2, 3,···), as a function of colony size. Measuring the fluctuations permits a validation of the conditions required for the application of a quantitative version of the Le Chatelier's principle, as derived using information theory. This principle provides a quantitative prediction of the role of perturbations and allows a characterization of a protein-protein interaction network.

Convergence of logic of cellular regulation in different premalignant cells by an information theoretic approach.

BACKGROUND: Surprisal analysis is a thermodynamic-like molecular level approach that identifies biological constraints that prevents the entropy from reaching its maximum. To examine the significance of altered gene expression levels in tumorigenesis we apply surprisal analysis to the WI-38 model through its precancerous states. The constraints identified by the analysis are transcription patterns underlying the process of transformation. Each pattern highlights the role of a group of genes that act coherently to define a transformed phenotype. RESULTS: We identify a major transcription pattern that represents a contraction of signaling networks accompanied by induction of cellular proliferation and protein metabolism, which is essential for full transformation. In addition, a more minor, "tumor signature" transcription pattern completes the transformation process. The variation with time of the importance of each transcription pattern is determined. Midway through the transformation, at the stage when cells switch from slow to fast growth rate, the major transcription pattern undergoes a total inversion of its weight while the more minor pattern does not contribute before that stage. CONCLUSIONS: A similar network reorganization occurs in two very different cellular transformation models: WI-38 and the cervical cancer HF1 models. Our results suggest that despite differences in a list of transcripts expressed in different cancer models the rationale of the network reorganization remains essentially the same.

All-DNA finite-state automata with finite memory.

Biomolecular logic devices can be applied for sensing and nano-medicine. We built three DNA tweezers that are activated by the inputs H(+)/OH(-); ; nucleic acid linker/complementary antilinker to yield a 16-states finite-state automaton. The outputs of the automata are the configuration of the respective tweezers (opened or closed) determined by observing fluorescence from a fluorophore/quencher pair at the end of the arms of the tweezers. The system exhibits a memory because each current state and output depend not only on the source configuration but also on past states and inputs.

Information-theoretic analysis of phenotype changes in early stages of carcinogenesis.

Cancer is a multistep process characterized by altered signal transduction, cell growth, and metabolism. To identify such processes in early carcinogenesis we use an information theoretic approach to characterize gene expression quantified as mRNA levels in primary keratinocytes (K) and human papillomavirus 16 (HPV16)-transformed keratinocytes (HF1 cells) from early (E) and late (L) passages and from benzo(a)pyrene-treated (BP) L cells. Our starting point is that biological signaling processes are subjected to the same quantitative laws as inanimate, nonequilibrium chemical systems. Environmental and genomic constraints thereby limit the maximal thermodynamic entropy that the biological system can reach. The procedure uncovers the changes in gene expression patterns in different networks and defines the significance of each altered network in the establishment of a particular phenotype. The development of transformed HF1 cells is shown to be represented by one major transcription pattern that is important at all times. Two minor transcription patterns are also identified, one that contributes at early times and a distinguishably different pattern that contributes at later times. All three transcription patterns defined by our analysis were validated by gene expression values and biochemical means. The major transcription pattern includes reduced transcripts participating in the apoptotic network and enhanced transcripts participating in cell cycle, glycolysis, and oxidative phosphorylation. The two minor patterns identify genes that are mainly involved in lipid or carbohydrate metabolism.

Maximal entropy inference of oncogenicity from phosphorylation signaling.

Point mutations in the phosphorylation domain of the Bcr-Abl fusion oncogene give rise to drug resistance in chronic myelogenous leukemia patients. These mutations alter kinase-mediated signaling function and phenotypic outcome. An information theoretic analysis of the correlation of phosphoproteomic profiling and transformation potency of the oncogene in different mutants is presented. The theory seeks to predict the leukemic transformation potency from the observed signaling by constructing a distribution of maximal entropy of site-specific phosphorylation events. The theory is developed with special reference to systems biology where high throughput measurements are typical. We seek sets of phosphorylation events most contributory to predicting the phenotype by determining the constraints on the signaling system. The relevance of a constraint is measured by how much it reduces the value of the entropy from its global maximum, where all events are equally likely. Application to experimental phospho-proteomics data for kinase inhibitor-resistant mutants shows that there is one dominant constraint and that other constraints are not relevant to a similar extent. This single constraint accounts for much of the correlation of phosphorylation events with the oncogenic potency and thereby usefully predicts the trends in the phenotypic output. An additional constraint possibly accounts for biological fine structure.

An electronic time scale in chemistry.

Ultrafast, subfemtosecond charge migration in small peptides is discussed on the basis of computational studies and compared with the selective bond dissociation after ionization as observed by Schlag and Weinkauf. The reported relaxation could be probed in real time if the removal of an electron could be achieved on the attosecond time scale. Then the mean field seen by an electron would be changing rapidly enough to initiate the migration. Tyrosine-terminated tetrapeptides have a particularly fast charge migration where in <1 fs the charge arrives at the other end. A femtosecond pulse can be used to observe the somewhat slower relaxation induced by correlation between electrons of different spins. A slower relaxation also is indicated when removing a deeper-lying valence electron. When a chromophoric amino acid is at one end of the peptide, the charge can migrate all along the peptide backbone up to the N end, but site-selective ionization is probably easier to detect for tryptophan than for tyrosine.

A phase I-II trial of induction chemotherapy with carboplatin and fluorouracil in locally advanced head and neck squamous cell carcinoma: a report from the UCL-Oncology Group, Belgium.

Eighty-three patients (median age, 56 years and Karnofsky performance status greater than or equal to 70) were treated with carboplatin (Carbo) and fluorouracil (5Fu) for stage III and IV head and neck squamous cell carcinoma (HNSCC). 5Fu (1 g/m2/d) was administered from day 1 to 4 by continuous infusion. Carbo was given on day 1 and, in order to evaluate its maximum-tolerated dose (MTD), the dose level was progressively increased from 250 mg/m2 to 450 mg/m2. The effectiveness of this association and its potential role in local control were also evaluated. Three patients received Carbo at a dose of 250 mg/m2, 13 received 300 mg/m2, one received 330 mg/m2, 12 received 350 mg/m2, six received 375 mg/m2, 26 received 400 mg/m2, 18 received 420 mg/m2, and four received 450 mg/m2. Two (13 of 83) or three courses (64 of 83), repeated every 4 weeks, were administered. The overall (primary tumor and node) response and complete response (CR) rates were 33% and 14%, respectively. For primary tumor, the response rate (RR) was 57% with 32% CR and 18% pathologic complete response (PCR); the RR was higher for patients with oropharyngeal tumor (76%, P = .037) and for patients treated with Carbo greater than or equal to 350 mg/m2 (65%, P = .02); the tumor size (T1 + T2 v T3 + T4) was a good prognostic factor for RR (90% v 46%, P = .001), CR (65% v 20%, P less than .001), and PCR (45% v 8%, P less than .001). For nodes, the RR was 33% with 11% CR. Grade 3-4 neutropenia and thrombocytopenia were experienced by 17% and 28% of the patients treated with 420 mg/m2 of Carbo and by 50% of the patients treated with 450 mg/m2. The MTD can be fixed at 420 mg/m2 and the proposed dose at 400 mg/m2. Thirty-eight patients were treated with surgery plus radiotherapy, 33 with radiotherapy alone, and seven with surgery alone. The median follow-up is 12 months. The 18-month disease-free survival (DFS) is 78% for overall complete responders and 39% for the others (P = .04). There is no primary tumor recurrence among the 12 patients with a primary tumor PCR treated by radiotherapy alone for tumor control (median follow-up, 17.3 months). The association of Carbo-5Fu is a safe induction chemotherapy regimen for HNSCC. The proposed dose of Carbo for future treatment is 400 mg/m2.(ABSTRACT TRUNCATED AT 400 WORDS)