Blockchain Papers

Sanket Motling, Satvik Razdan, Sankaranarayanan Sengunther

No abstract is available for this record.

Rudi Agius, Anders C. Riis-Jensen, Bettina Wimmer, Caspar da Cunha‐Bang · 11 authors

No abstract is available for this record.

Stefan W. Krause

In this issue of Cytometry A, Zhao et al. (page 1073–1080) report on their work to diagnose leukemic B cell non-Hodgkin's Lymphoma from flow cytometry (FCM) raw data of blood and bone marrow samples using a dedicated computer approach, which would assign one of eight B-cell lymphoma diagnoses or “normal” to a sample. A remarkable of level of classification performance could be achieved in the validation set. For the “true” classification of B-cell lymphomas, conventional diagnostics had incorporated morphology, FCM and additional information from histology and genetics if needed, whereas computer diagnosis was derived from FCM data alone. In this context, uncertainty to delineate, for example, monoclonal B-cell lymphocytosis from chronic lymphocytic leukemia or to subclassify a B-cell malignancy as either mantle cell lymphoma or prolymphocytic leukemia is not an outright error, but is rather based on the limitations of FCM itself. Furthermore, cell populations tagged as abnormal by the algorithm and color-coded accordingly in conventional plots can help human diagnosticians to review and fine-tune the diagnosis. However, some lymphomas (most prominent in follicular lymphoma) were classified as normal by the algorithm. Vice versa, only few samples classified as “normal” by human diagnosticians were classified as lymphoma by the algorithm. Thus, a deficit in sensitivity exists, which is clinically relevant. Computer support is instrumental for the analysis of FCM data, because nobody is able to draw conclusions from raw list mode files. However, conventional FCM computer programs execute relative simple tasks to support the workflow of a human researcher or diagnostician. In a typical workflow, several sequential steps have to be performed (Fig. 1, left side). Fluorescence spillover compensation is calculated from control samples. One-dimensional transformation of raw data (logarithmic, logical, possibly a shift of zero and negative values to some defined minimum, etc.) is routinely performed on fluorescence channels. Data are displayed in histograms or two-dimensional plots. Starting gates are used to look for artifacts and to remove debris and cells not of interest. A considerable number of plots are necessary, if several fluorochromes are used and several populations are of interest. Data from several samples with identical panel may be displayed in parallel in an overlay. Cells are tagged according to gates in these plots and may then be displayed separately and/or color-coded. Hierarchical and/or Boolean gating strategies are used for the definition of cell populations and subpopulations of interest. Cell numbers and antigen expression of these cell populations of interest constitute the readout of a single tube. A final result or diagnosis is derived assessing this readout or the synopsis of the readout of several tubes. All of the calculations in such a manual workflow are based on straight “if A then B” logic, performing calculations on a maximum of two parameters concurrently. Conventional FCM computer support aims at displaying data in a clear manner to the human operator, especially effects of manipulation in two-parameter plots upon plots of other parameters, but not at automation. The most advanced process in standard applications is the calculation of fluorescence spillover compensation, which nowadays usually is performed in some (semi-) automated fashion. However, although every single step in this procedure is quite straightforward, due to the multitude of plots and gates from current 10 to 14 parameter FCM data, important information may be missed. In the recent decades, many attempts have been reported to introduce more advanced computation methods into histology, cytopathology, image cytometry and conventional FCM analysis (1, 2). These algorithms will be called artificial intelligence (AI) from here, although some of them do not deserve this name in its strict sense. Two strategies, sometimes overlapping, are applied in these attempts: firstly, AI may be used to automatize conventional data processing and analysis as described above in order to reduce the workload for the investigator, reduce bias using standardized procedures, and speed up analyses. To this end, regarding FCM, algorithms search for minimal values in distributions to define optimal positions for gates to divide populations or search for appropriate cut off values to gate out debris. Furthermore, normalization algorithms can be applied to level out differences due to instrument settings or biological variations in sets of multiple similar data. Many of these algorithms are available in the Bioconductor “flow Core” FCM package implemented in R (3). Secondly, new methods were introduced that go beyond the sequential analysis of two-dimensional plots and base calculations on more parameters of the higher-dimensional space in parallel, which is a crucial need nowadays, when standard cytometers report 10 to 14 parameters per cell and dedicated research instruments up to over 100 parameters. Such algorithms can either substitute conventional strategies, for example, to gate cell populations and read out antigen expression levels or they can be used to extract information from the raw data that is not accessible by conventional gating (4). One of the prominent tasks within an FCM workflow is to define cell populations within a mixture of different cells (“clustering”) that may be of interest for research or diagnosis. AI can directly use higher dimensional data as input for cell clustering or it can perform dimensionality reduction and data visualization, for example, by tSNE or one of its variants (5, 6) or SOM (7), the latter already including some clustering of the data. After dimensionality reduction, population clustering can be added by separate AI algorithms or a human operator can take over for this task, integrating the output of the dimensionality reduction and conventional gating. Many different algorithms are able to solve the task of clustering in an automated fashion either performing a two-step procedure integrating dimension reduction and subsequential clustering or direct clustering of higher dimensional data; however, as shown in the FlowCAP challenges, results are not unequivocal, especially, if the number of clusters is not defined a priori, and differences remain between different algorithms and human experts. Up to now, no perfect automatic solution for cell clustering exists, although many solutions perform quite well (7). Furthermore, clustering revealing further information on relatedness between populations has been suggested for a multitude of different research questions, for example, cellular developmental trajectories, and has been optimized according to these special tasks (further Ref. in 4). Furthermore, metadata extracted from raw FCM data may also be clustered, for example, in order to define diagnostic or prognostic subgroups (8). Whereas unsupervised clustering can be helpful for many exploratory research questions to identify cell populations and subpopulations, for medical diagnostic purposes supervised AI methods have been described, that use external information such as diagnoses or outcome to train the AI, for example, using support vector machines or neural networks. All of these strategies rely on a large dataset for training and may incorporate more or less steps from a conventional workflow (4, 9, 10). Manual gating and tagging of cell populations may be used for training of the AI (11) or AI may be trained using only the final results, that is, diagnosis, as described, for example, in Ref. (12) or in the work by Zhao et al. discussed here. Several AI strategies have been able to discern overt acute myeloid leukemia (AML) from normal samples with a high success rate in the second FlowCap challenge (7), however, this can be a considered a quite simple task, since overt AML is easily characterized by a large abnormal population of blast or sometimes monocytic cells. In contrast, separation of AML from myelodysplastic syndromes or from acute lymphoblastic leukemia, everyday questions in diagnostics, is less trivial. In contrast to simplified “yes or no” tasks, Zhao et al. tackled a much more realistic question: to deduce a specific diagnosis from FCM panels as they are used in conventional diagnostics. They achieved this goal without an attempt to mimic a conventional human FCM workflow. They transformed the FCM data by self-organizing maps (SOM) and classified these representations by a convolutional neural network (CNN), dealing with each tube separately first and finally with data from all three tubes. The researchers took advantage of a very large database of patient sample FCM data. Data from more than 18,000 samples analyzed in a uniform fashion with identical antibody combinations and more than 200 samples of the rarest subtype of lymphoma could be used to train the CNN. In order to get some insight into the CNN “black box,” they checked, which markers were of most importance for the AI to classify a specific diagnosis correctly and they had cell populations tagged that were detected to be abnormal and discriminative by the algorithm for the respective disease in a way to understand the AI's decision (and to use this assignment for a possible refinement by a human diagnostician in practical diagnostic use in the future). As described above, the results of their approach are remarkable, but a problem in sensitivity to detect all true lymphoma cases remains, which is most prominent for follicular lymphoma. Maybe the CNN could be trained in a way, that the correct distinction B-NHL of any type versus normal is assigned a higher weight compared to B-NHL subtyping. If we inspect the importance of single markers for AI performance in Supporting Figure 5, we note that some diagnosis assignments rely heavily on a few markers, whereas other diagnoses seem to rather depend on the distribution of many markers. Interestingly, the latter diagnoses without dependence on dominant markers have the highest rate of falsely being categorized as normal (follicular lymphoma, marginal zone lymphoma, lymphoplasmactic lymphoma). Furthermore, for a human diagnostician, an imbalance of kappa versus lambda light chain expression on B cells is a very important clue for a diagnosis of B-cell lymphoma, whereas the CNN of Zhao et al. does not seem to rely heavily on this information. In a different approach, to detect minimal residual disease in childhood acute leukemia, conventional gating was used to train a machine learning algorithm based on Gaussian mixture models (11). Thus, for the non-AI expert the idea comes up, if some information of a conventional workflow, collected by an automated application, could be “injected” into a CNN algorithm. If we assume that the problem of sensitivity will be tackled by improved versions in the near future, the AI solution of Zhao et al. will in fact be able to perform at “hematologist-level” and may even deliver B-NHL subtyping competence exceeding the results of conventional FCM alone. However, further problems have to be solved for a broader uptake of such a method: different laboratories work with different antibody panels and even antibodies recognizing the same cluster of differentiation antigen behave differently due to different antibody clones, different fluorochromes and different spillover from other fluorochromes in the panel. Thus, some methods of knowledge transfer are needed, if we want to avoid starting again with a training sample of more than 10,000 cases for every new antibody panel. If researchers will be able to solve these problems, AI for diagnostic FCM may finally leave the “proof of concept” stage and enter routine diagnostics. Open access funding enabled and organized by Projekt DEAL.

C. G. Schmidt

This is part of an occasional series that recalls some of the stories reported 10 years ago in the News section of the Journal. In 1998, JNCI published a two-part feature about efforts to overcome multidrug resistance in cancer, which at the time was thought to be governed chiefly by a superfamily of molecular transporters, known as ATP-binding cassette (ABC) proteins. ABC transporters act as efflux pumps, which expel toxins and drugs from a cell. By reversing those transporters in cancer cells—particularly P-glycoprotein (Pgp), which was, and still is, the best-characterized among them—researchers hoped to overcome drug resistance, a primary cause of treatment failure. A decade later, efforts to reverse Pgp have proven futile, in part because the transport protein is also expressed by healthy tissues—leading to unacceptable side effects when its activity is knocked out. Furthermore, dozens of other ABC transport proteins identified within the last 10 years can take over for Pgp when its activity is reversed in cancer cells. Scientists are still studying clinical opportunities with ABC transporters, but drug resistance in cancer remains as much a problem now as it ever was. Even so, scientists have new leads to follow. Advances in molecular biology, driven by genomics and related fields, have revealed new resistance mechanisms and broadened opportunities to overcome it, said Michael M. Gottesman , M.D., head of the National Cancer Institute's molecular cell genetics section. “We just have to apply the mechanistic knowledge we’re gaining in the laboratory towards progress in the clinic.” Michael M. Gottesman, M.D. Today, those efforts are proceeding on dual fronts. Whereas scientists 10 years ago were concerned only with multidrug resistance against chemotherapy, those working today must also contend with single-drug resistance to newer, targeted therapies such as imatinib (Gleevec), a drug for chronic myelogenous leukemia (CML) and other cancers that inactivates a cancer-inducing protein called Bcr-Abl. Unlike chemotherapy—which kills rapidly dividing cells regardless of whether they’re cancerous—targeted therapies interfere with specific molecules involved in cancer and tumor growth. The new era of targeted therapy was supposed to leave chemotherapy behind and multidrug resistance behind with it. That's because researchers associated such resistance almost exclusively with efflux pumps, which act against a broad array of natural and synthetic toxins. By using nontoxic compounds targeted against specific molecules in carcinogenesis, scientists hoped that they could avoid resistance. But those expectations were dashed when clinicians found that although patients typically respond well to imatinib—among the first targeted drugs to reach the market—many also relapse within 3–5 years. Confronted with that distressing problem, scientists were reluctant to blame mechanisms like efflux, which were generally attributed to multidrug resistance and chemotherapy, for imatinib resistance. Instead, their explanations veered toward gene mutations that, by chance, might allow CML cells to survive imatinib exposure. In that scenario, most CML cells are killed by the drug, but those with mutations that favor survival multiply over time, leading to relapse. Scientists call this type of drug resistance “acquired.” To an extent, subsequent research has borne out this hypothesis, which was proposed when imatinib resistance was first observed, within a few years of the drug's approval by the U.S. Food and Drug Administration in 2001. More than a dozen gene mutations have since been implicated, of which the most effective is T315I, according to Susan Bates, M.D., director of the NCI's molecular therapeutics section. This simple variation alters Bcr-Abl's three-dimensional binding site, thus blocking imatinib's access to the target protein. But Bates argues that mutations alone can’t account entirely for imatinib resistance. Patients whose leukemic cells have identical mutations can vary with respect to the degree of resistance, she said, which suggests that other factors—including mechanisms linked to multidrug resistance—are also at play. “For example, interindividual variation has been found in the amount of Gleevec that winds up in the bloodstream, which could reflect differences in cellular absorption of the drug,” Bates explained. “We can’t rule out that ABC transporters might be involved, given that Gleevec is a substrate for at least two: Pgp and breast cancer resistance protein. In CML cells, those transporters reduce the amount of Gleevec that reaches the protein target, and that fosters resistance.” (However, variation in resistance among patients could also be due in part to behavioral factors, such as adherence to the drug regimen [ see JNCI 2008; 100: 912–3 ].) While researchers work to tease out resistance mechanisms against targeted therapies, efforts to overcome multidrug resistance to chemotherapy are making headway. Researchers in Gottesman's lab at NCI, who are among the leaders in this area, approach the problem by using three general strategies. Gottesman's research team collaborates with pharmaceutical companies to develop drugs that circumvent known resistance mechanisms. Compounds that evade ABC transporters—of which there are 48, according to current estimates—fall into that general category. Epothilones, for example, a new class of cytotoxic molecules identified as potential chemotherapy agents, aren’t recognized by Pgp. “This provides proof of concept that new classes of anticancer agents that don’t interact with multidrug transporters can be developed,” he said. This approach is currently in early developmental stages, with no compounds in clinical trials. Applying an alternate approach, the team strives to block resistance by inhibiting or reversing it. Clinical applications based on this approach would involve drug combinations: one drug to inhibit the resistance mechanism and another to kill the cancer cell. This strategy already has a long and unsuccessful history, exemplified by compounds such as PSC-833, a Pgp-reversing agent tested often in clinical trials. PSC-833 failed in part because of unpredictable pharmacokinetic interactions, leading to underdosing in some patients and overdosing in others. Scientists now hope to overcome that problem with third- and fourth-generation inhibitors designed for low pharmacokinetic interactions, as well as tight binding with target proteins. According to researchers in Gottesman's laboratory, promising candidates include tariquidar, a compound that offers extended Pgp inhibition, now in phase III clinical trials, and CBT-1, a plant alkaloid that inhibits Pgp in addition to an ABC transporter known as MRP1. The third strategy exploits certain unique features of multidrug-resistant cells, such as high surface expression of Pgp or a paradoxical hypersensitivity to a range of compounds. These features make it possible to target these cells directly, Gottesman said. For instance, thiosemicarbazones—a class of compounds with known antiviral, antimicrobial, and anticancer activity—kill resistant cells through Pgp-related mechanisms. This approach has not yet produced compounds for clinical trials. Scientists have also focused on the influx mechanisms that cells use to absorb certain compounds, including cancer drugs, instead of the efflux pumps that they use to expel them. Solute-carrier (SLC) protein transporters, recognized as the largest superfamily of membrane proteins, participate in these influx processes. Jeffrey Moscow, M.D., chief of pediatric hematology/oncology at the University of Kentucky Medical Center in Lexington, is now working to identify SLCs expressed uniquely by cancer cells. By harnessing these proteins, he hopes to pump drugs into a cancer cell faster than efflux transporters can pump them out. He's already identified an SLC expressed uniquely by lung and gastrointestinal malignancies, known as OATBIB3, and another expressed by leukemia, known as SLC22A16. “In these cases, the specificity of therapy would be determined by the expression of the solute carriers,” he said. Efforts to overcome single-drug resistance to targeted therapies differ from those applied to multidrug resistance. Instead of circumventing, inhibiting, or targeting mechanisms that cancer cells use to avoid poisons, scientists try to augment their treatment options with additional molecular targets, said Jeffrey Settleman, Ph.D., a professor at Harvard Medical School and scientific director of the Massachusetts General Hospital Cancer Center. Settleman's laboratory has accumulated the largest collection of human cancer cell lines in existence—more than 1,000 in all, representing all the major tissue types. With automated screening technology, he and his colleagues test putative targeted therapies until they find a positive hit in a sensitive cell line. Those lines are treated with the drug until only the resistant cells remain. By culturing those cells, Settleman's team can look for mutations that confer resistance, as well as for new targets to which the cells might be sensitive. The clinical strategy for managing resistance in targeted therapy, Settleman said, entails drug combinations to inhibit several targets sequentially. “The scenario could be that we convert cancer to a chronic disease by giving drugs in succession; as resistance develops to one drug, we shift to another,” he said. This approach is already being used now. For instance, CML patients who become resistant to imatinib can be treated with dasatinib (Sprycel). In a melanoma cell line, Settleman's laboratory recently modeled the acquisition of resistance to a candidate Raf kinase inhibitor. They found that resistant cells merely switched their metabolic dependency from Raf to another related kinase. “Raf kinase inhibitors are being tested clinically now,” Settleman said. “It's early days for these compounds, and we’re not sure they’re going to work. But if they do, we may have zeroed in on the resistance mechanism in advance.” The ability to switch from one kinase to another shows how adaptable cancer cells are in the face of drug pressure. But fortunately, it appears that cancer cells have at most three to four resistance mechanisms that they can direct against any particular compound, Settleman said. “And that shows we’re ultimately dealing with a manageable problem,” he said. “It indicates how important it is to tackle resistance from more than one angle. We’re faced with the same resistance mechanisms that infectious microbes use to evolve and mutate their way around a drug. We treat [human immunodeficiency virus] with multidrug cocktails, and we’ll be doing more of the same with cancer; our aim is to cut off the cells’ options to adapt.” So, although the previous 10 years revealed new mechanisms in cancer cell biology, including roles played by cancer stem cells that could offer the best therapeutic targets of all, accelerated research during the coming decade might bring the problem of resistance under better control. But doing that won’t be easy, Settleman admitted. “Cancer cells are crafty organisms in their own right,” he said. “And resistance is still the final frontier in treatment.”