Early Glaucoma Detection by Using Style Transfer to Predict Retinal Nerve Fiber Layer Thickness Distribution on the Fundus Photograph.

Objective: We aimed to develop a deep learning (DL)-based algorithm for early glaucoma detection based on color fundus photographs that provides information on defects in the retinal nerve fiber layer (RNFL) and its thickness from the mapping and translating relations of spectral domain OCT (SD-OCT) thickness maps. Design: Developing and evaluating an artificial intelligence detection tool. Subjects: Pretraining paired data of color fundus photographs and SD-OCT images from 189 healthy participants and 371 patients with early glaucoma were used. Methods: The variational autoencoder (VAE) network training architecture was used for training, and the correlation between the fundus photographs and RNFL thickness distribution was determined through the deep neural network. The reference standard was defined as a vertical cup-to-disc ratio of ≥0.7, other typical changes in glaucomatous optic neuropathy, and RNFL defects. Convergence indicates that the VAE has learned a distribution that would enable us to produce corresponding synthetic OCT scans. Main Outcome Measures: Similarly to wide-field OCT scanning, the proposed model can extract the results of RNFL thickness analysis. The structural similarity index measure (SSIM) and peak signal-to-noise ratio (PSNR) were used to assess signal strength and the similarity in the structure of the color fundus images converted to an RNFL thickness distribution model. The differences between the model-generated images and original images were quantified. Results: We developed and validated a novel DL-based algorithm to extract thickness information from the color space of fundus images similarly to that from OCT images and to use this information to regenerate RNFL thickness distribution images. The generated thickness map was sufficient for clinical glaucoma detection, and the generated images were similar to ground truth (PSNR: 19.31 decibels; SSIM: 0.44). The inference results were similar to the OCT-generated original images in terms of the ability to predict RNFL thickness distribution. Conclusions: The proposed technique may aid clinicians in early glaucoma detection, especially when only color fundus photographs are available.

Necessity of Local Modification for Deep Learning Algorithms to Predict Diabetic Retinopathy.

Deep learning (DL) algorithms are used to diagnose diabetic retinopathy (DR). However, most of these algorithms have been trained using global data or data from patients of a single region. Using different model architectures (e.g., Inception-v3, ResNet101, and DenseNet121), we assessed the necessity of modifying the algorithms for universal society screening. We used the open-source dataset from the Kaggle Diabetic Retinopathy Detection competition to develop a model for the detection of DR severity. We used a local dataset from Taipei City Hospital to verify the necessity of model localization and validated the three aforementioned models with local datasets. The experimental results revealed that Inception-v3 outperformed ResNet101 and DenseNet121 with a foreign global dataset, whereas DenseNet121 outperformed Inception-v3 and ResNet101 with the local dataset. The quadratic weighted kappa score (κ) was used to evaluate the model performance. All models had 5-8% higher κ for the local dataset than for the foreign dataset. Confusion matrix analysis revealed that, compared with the local ophthalmologists' diagnoses, the severity predicted by the three models was overestimated. Thus, DL algorithms using artificial intelligence based on global data must be locally modified to ensure the applicability of a well-trained model to make diagnoses in clinical environments.

Axitinib in combination with radiotherapy for advanced hepatocellular carcinoma: a phase I clinical trial.

BACKGROUND: To investigate maximum tolerated dose (MTD) of axitinib, a selective vascular endothelial growth factor receptor 1-3 inhibitor, in combination with radiotherapy (RT) for advanced hepatocellular carcinoma (HCC). METHODS: This phase I study followed the rule of traditional 3 + 3 design. Major eligibility included: (1) patients with advanced HCC unsuitable for surgery, radiofrequency ablation or transarterial chemoembolization, or who failed after prior local-regional treatment; (2) failure on sorafenib or no grant for sorafenib from health insurance system. Eligible patients with advanced HCC received axitinib for total 8 weeks during and after RT. Three cohorts with axitinib dose escalation were planned: 1 mg twice daily (level I), 2 mg twice daily (level II) and 3 mg twice daily (level III). The prescribed doses of RT ranged from 37.5 to 67.5 Gy in 15 fractions to liver tumor(s) and were determined based on an upper limit of mean liver dose of 18 Gy (intended isotoxic RT for normal liver). The primary endpoint was MTD of axitinib in combination with RT. The secondary endpoints included overall response rate (ORR), RT in-field response rate, acute and late toxicities, overall survival (OS) and progression free survival (PFS). RESULTS: Total nine eligible patients received axitinib dose levels of 1 mg twice daily (n = 3), 2 mg twice daily (n = 3) and 3 mg twice daily (n = 3). Dose-limiting toxicity (DLT) did not occur in the 3 cohorts; the MTD was defined as 3 mg twice daily in this study. ORR was 66.7%, including 3 complete responses and 3 partial responses, at 3 months after treatment initiation. With a median follow-up of 16.6 months, median OS was not reached, 1-year OS was 66.7%, and median PFS was 7.4 months. CONCLUSIONS: Axitinib in combination with RT for advanced HCC was well tolerated with an axitinib MTD of 3 mg twice daily in this study. The outcome analysis should be interpreted with caution due to the small total cohort. Trial registration ClinicalTrials.gov (Identifier: NCT02814461), Registered June 27, 2016-Retrospectively registered, https://clinicaltrials.gov/ct2/show/NCT02814461.

A Pliable Mediator Acts as a Functional Rather Than an Architectural Bridge between Promoters and Enhancers.

While Mediator plays a key role in eukaryotic transcription, little is known about its mechanism of action. This study combines CRISPR-Cas9 genetic screens, degron assays, Hi-C, and cryoelectron microscopy (cryo-EM) to dissect the function and structure of mammalian Mediator (mMED). Deletion analyses in B, T, and embryonic stem cells (ESC) identified a core of essential subunits required for Pol II recruitment genome-wide. Conversely, loss of non-essential subunits mostly affects promoters linked to multiple enhancers. Contrary to current models, however, mMED and Pol II are dispensable to physically tether regulatory DNA, a topological activity requiring architectural proteins. Cryo-EM analysis revealed a conserved core, with non-essential subunits increasing structural complexity of the tail module, a primary transcription factor target. Changes in tail structure markedly increase Pol II and kinase module interactions. We propose that Mediator's structural pliability enables it to integrate and transmit regulatory signals and act as a functional, rather than an architectural bridge, between promoters and enhancers.

Genetic determinants of co-accessible chromatin regions in activated T cells across humans.

Over 90% of genetic variants associated with complex human traits map to non-coding regions, but little is understood about how they modulate gene regulation in health and disease. One possible mechanism is that genetic variants affect the activity of one or more cis-regulatory elements leading to gene expression variation in specific cell types. To identify such cases, we analyzed ATAC-seq and RNA-seq profiles from stimulated primary CD4+ T cells in up to 105 healthy donors. We found that regions of accessible chromatin (ATAC-peaks) are co-accessible at kilobase and megabase resolution, consistent with the three-dimensional chromatin organization measured by in situ Hi-C in T cells. Fifteen percent of genetic variants located within ATAC-peaks affected the accessibility of the corresponding peak (local-ATAC-QTLs). Local-ATAC-QTLs have the largest effects on co-accessible peaks, are associated with gene expression and are enriched for autoimmune disease variants. Our results provide insights into how natural genetic variants modulate cis-regulatory elements, in isolation or in concert, to influence gene expression.

The Energetics and Physiological Impact of Cohesin Extrusion.

Cohesin extrusion is thought to play a central role in establishing the architecture of mammalian genomes. However, extrusion has not been visualized in vivo, and thus, its functional impact and energetics are unknown. Using ultra-deep Hi-C, we show that loop domains form by a process that requires cohesin ATPases. Once formed, however, loops and compartments are maintained for hours without energy input. Strikingly, without ATP, we observe the emergence of hundreds of CTCF-independent loops that link regulatory DNA. We also identify architectural "stripes," where a loop anchor interacts with entire domains at high frequency. Stripes often tether super-enhancers to cognate promoters, and in B cells, they facilitate Igh transcription and recombination. Stripe anchors represent major hotspots for topoisomerase-mediated lesions, which promote chromosomal translocations and cancer. In plasmacytomas, stripes can deregulate Igh-translocated oncogenes. We propose that higher organisms have coopted cohesin extrusion to enhance transcription and recombination, with implications for tumor development.

Comparing the Effectiveness of Combined External Beam Radiation and Hyperthermia Versus External Beam Radiation Alone in Treating Patients With Painful Bony Metastases: A Phase 3 Prospective, Randomized, Controlled Trial.

PURPOSE: To compare the response, duration of pain relief, and time to achieve complete pain relief after radiation therapy (RT) with or without hyperthermia (HT) in patients with painful bony metastases. METHODS AND MATERIALS: Cancer patients with bony metastases and pain score ≥4 on the Brief Pain Inventory (BPI) were randomized to RT of 30 Gy in 10 fractions combined with HT (RT + HT) versus RT alone. Hyperthermia was performed by the Thermotron RF-8, with maintenance of the target temperature for 40 minutes per treatment within 2 hours after RT, twice weekly for 2 weeks. Patients were stratified by lesion number (solitary or multiple), BPI score (4-6 vs 7-10), and primary site. The primary endpoint was complete response (CR) (BPI = 0 with no increase of analgesics) within 3 months after treatment. This study was registered with ClinicalTrials.gov. RESULTS: The study was terminated early after an interim analysis of 57 patients, 3 years after the first enrollment (November 2013 to November 2016): 29 patients in the RT + HT group and 28 patients in the RT-alone group. The CR rate at 3 months after treatment was 37.9% in the RT + HT group versus 7.1% in the RT-alone group (P=.006). The accumulated CR rate within 3 months after treatment was 58.6% in the RT + HT group versus 32.1% in the RT-alone group (P=.045). Median time to pain progression was 55 days in patients with CR (n=9) in the RT-alone group, whereas the endpoint was not reached during the 24-week follow-up in the RT + HT group (P<.01). CONCLUSIONS: The addition of HT to RT significantly increases the pain control rate and extends response duration compared with RT alone for painful bony metastases.

Cohesin Loss Eliminates All Loop Domains.

The human genome folds to create thousands of intervals, called "contact domains," that exhibit enhanced contact frequency within themselves. "Loop domains" form because of tethering between two loci-almost always bound by CTCF and cohesin-lying on the same chromosome. "Compartment domains" form when genomic intervals with similar histone marks co-segregate. Here, we explore the effects of degrading cohesin. All loop domains are eliminated, but neither compartment domains nor histone marks are affected. Loss of loop domains does not lead to widespread ectopic gene activation but does affect a significant minority of active genes. In particular, cohesin loss causes superenhancers to co-localize, forming hundreds of links within and across chromosomes and affecting the regulation of nearby genes. We then restore cohesin and monitor the re-formation of each loop. Although re-formation rates vary greatly, many megabase-sized loops recovered in under an hour, consistent with a model where loop extrusion is rapid.

Myc Regulates Chromatin Decompaction and Nuclear Architecture during B Cell Activation.

50 years ago, Vincent Allfrey and colleagues discovered that lymphocyte activation triggers massive acetylation of chromatin. However, the molecular mechanisms driving epigenetic accessibility are still unknown. We here show that stimulated lymphocytes decondense chromatin by three differentially regulated steps. First, chromatin is repositioned away from the nuclear periphery in response to global acetylation. Second, histone nanodomain clusters decompact into mononucleosome fibers through a mechanism that requires Myc and continual energy input. Single-molecule imaging shows that this step lowers transcription factor residence time and non-specific collisions during sampling for DNA targets. Third, chromatin interactions shift from long range to predominantly short range, and CTCF-mediated loops and contact domains double in numbers. This architectural change facilitates cognate promoter-enhancer contacts and also requires Myc and continual ATP production. Our results thus define the nature and transcriptional impact of chromatin decondensation and reveal an unexpected role for Myc in the establishment of nuclear topology in mammalian cells.

Genome Organization Drives Chromosome Fragility.

In this study, we show that evolutionarily conserved chromosome loop anchors bound by CCCTC-binding factor (CTCF) and cohesin are vulnerable to DNA double strand breaks (DSBs) mediated by topoisomerase 2B (TOP2B). Polymorphisms in the genome that redistribute CTCF/cohesin occupancy rewire DNA cleavage sites to novel loop anchors. While transcription- and replication-coupled genomic rearrangements have been well documented, we demonstrate that DSBs formed at loop anchors are largely transcription-, replication-, and cell-type-independent. DSBs are continuously formed throughout interphase, are enriched on both sides of strong topological domain borders, and frequently occur at breakpoint clusters commonly translocated in cancer. Thus, loop anchors serve as fragile sites that generate DSBs and chromosomal rearrangements. VIDEO ABSTRACT.

Deletion of DXZ4 on the human inactive X chromosome alters higher-order genome architecture.

During interphase, the inactive X chromosome (Xi) is largely transcriptionally silent and adopts an unusual 3D configuration known as the "Barr body." Despite the importance of X chromosome inactivation, little is known about this 3D conformation. We recently showed that in humans the Xi chromosome exhibits three structural features, two of which are not shared by other chromosomes. First, like the chromosomes of many species, Xi forms compartments. Second, Xi is partitioned into two huge intervals, called "superdomains," such that pairs of loci in the same superdomain tend to colocalize. The boundary between the superdomains lies near DXZ4, a macrosatellite repeat whose Xi allele extensively binds the protein CCCTC-binding factor. Third, Xi exhibits extremely large loops, up to 77 megabases long, called "superloops." DXZ4 lies at the anchor of several superloops. Here, we combine 3D mapping, microscopy, and genome editing to study the structure of Xi, focusing on the role of DXZ4 We show that superloops and superdomains are conserved across eutherian mammals. By analyzing ligation events involving three or more loci, we demonstrate that DXZ4 and other superloop anchors tend to colocate simultaneously. Finally, we show that deleting DXZ4 on Xi leads to the disappearance of superdomains and superloops, changes in compartmentalization patterns, and changes in the distribution of chromatin marks. Thus, DXZ4 is essential for proper Xi packaging.

Chromatin extrusion explains key features of loop and domain formation in wild-type and engineered genomes.

We recently used in situ Hi-C to create kilobase-resolution 3D maps of mammalian genomes. Here, we combine these maps with new Hi-C, microscopy, and genome-editing experiments to study the physical structure of chromatin fibers, domains, and loops. We find that the observed contact domains are inconsistent with the equilibrium state for an ordinary condensed polymer. Combining Hi-C data and novel mathematical theorems, we show that contact domains are also not consistent with a fractal globule. Instead, we use physical simulations to study two models of genome folding. In one, intermonomer attraction during polymer condensation leads to formation of an anisotropic "tension globule." In the other, CCCTC-binding factor (CTCF) and cohesin act together to extrude unknotted loops during interphase. Both models are consistent with the observed contact domains and with the observation that contact domains tend to form inside loops. However, the extrusion model explains a far wider array of observations, such as why loops tend not to overlap and why the CTCF-binding motifs at pairs of loop anchors lie in the convergent orientation. Finally, we perform 13 genome-editing experiments examining the effect of altering CTCF-binding sites on chromatin folding. The convergent rule correctly predicts the affected loops in every case. Moreover, the extrusion model accurately predicts in silico the 3D maps resulting from each experiment using only the location of CTCF-binding sites in the WT. Thus, we show that it is possible to disrupt, restore, and move loops and domains using targeted mutations as small as a single base pair.

Double autophagy modulators reduce 2-deoxyglucose uptake in sarcoma patients.

RATIONALE: According to the metabolic symbiosis model, cancer stromal fibroblasts could be hijacked by surrounding cancer cells into a state of autophagy with aerobic glycolysis to help provide recycled nutrients. The purpose of this study was to investigate whether combined treatment with the autophagy inhibitor: hydroxychloroquine (HCQ) and the autophagy inducer: sirolimus (rapamycin, Rapa) would reduce glucose utilization in sarcoma patients. METHODS: Ten sarcoma patients who failed first-line treatment were enrolled in this study. They were treated with 1 mg of Rapa and 200 mg of HCQ twice daily for two weeks. The standardized uptake values (SUV) from pretreatment and posttreatment [18F]-fluorodeoxyglucose positron emission tomography (FDG PET) scans were reviewed, and changes from the baseline SUVmax were evaluated. RESULTS: Based on FDG PET response criteria, six patients had a partial response; three had stable disease, and one had progressive disease. Nevertheless, none of them showed a reduction in tumor volume. The mean SUVmax reduction in the 34 lesions evaluated was - 19.6% (95% CI = -30.1% to -9.1%), while the mean volume change was +16.4% (95% CI = +5.8% to + 27%). Only grade 1 toxicities were observed. Elevated serum levels of lactate dehydrogenase were detected after treatment in most metabolic responders. CONCLUSIONS: The results of reduced SUVmax without tumor volume reduction after two weeks of Rapa and HCQ treatment may indicate that non-proliferative glycolysis occurred mainly in the cancer associated fibroblast compartment, and decreased glycolytic activity was evident from Rapa + HCQ double autophagy modulator treatment.

A novel dual kinase function of the RET proto-oncogene negatively regulates activating transcription factor 4-mediated apoptosis.

The RET proto-oncogene, a tyrosine kinase receptor, is widely known for its essential role in cell survival. Germ line missense mutations, which give rise to constitutively active oncogenic RET, were found to cause multiple endocrine neoplasia type 2, a dominant inherited cancer syndrome that affects neuroendocrine organs. However, the mechanisms by which RET promotes cell survival and prevents cell death remain elusive. We demonstrate that in addition to cytoplasmic localization, RET is localized in the nucleus and functions as a tyrosine-threonine dual specificity kinase. Knockdown of RET by shRNA in medullary thyroid cancer-derived cells stimulated expression of activating transcription factor 4 (ATF4), a master transcription factor for stress-induced apoptosis, through activation of its target proapoptotic genes NOXA and PUMA. RET knockdown also increased sensitivity to cisplatin-induced apoptosis. We observed that RET physically interacted with and phosphorylated ATF4 at tyrosine and threonine residues. Indeed, RET kinase activity was required to inhibit the ATF4-dependent activation of the NOXA gene because the site-specific substitution mutations that block threonine phosphorylation increased ATF4 stability and activated its targets NOXA and PUMA. Moreover, chromatin immunoprecipitation assays revealed that ATF4 occupancy increased at the NOXA promoter in TT cells treated with tyrosine kinase inhibitors or the ATF4 inducer eeyarestatin as well as in RET-depleted TT cells. Together these findings reveal RET as a novel dual kinase with nuclear localization and provide mechanisms by which RET represses the proapoptotic genes through direct interaction with and phosphorylation-dependent inactivation of ATF4 during the pathogenesis of medullary thyroid cancer.

Investigating electronic word-of-mouth effects on online discussion forums: the role of perceived positive electronic word-of-mouth review credibility.

Electronic word of mouth (eWOM) has been an important factor influencing consumer purchase decisions. Using the ABC model of attitude, this study proposes a model to explain how eWOM affects online discussion forums. Specifically, we propose that platform (Web site reputation and source credibility) and customer (obtaining buying-related information and social orientation through information) factors influence purchase intentions via perceived positive eWOM review credibility, as well as product and Web site attitudes in an online community context. A total of 353 online discussion forum users in an online community (Fashion Guide) in Taiwan were recruited, and structural equation modeling (SEM) was used to test the research hypotheses. The results indicate that Web site reputation, source credibility, obtaining buying-related information, and social orientation through information positively influence perceived positive eWOM review credibility. In turn, perceived positive eWOM review credibility directly influences purchase intentions and also indirectly influences purchase intentions via product and Web site attitudes. Finally, we discuss the theoretical and managerial implications of the findings.

Reciprocal complementation of the tumoricidal effects of radiation and natural killer cells.

The tumor microenvironment is a key determinant for radio-responsiveness. Immune cells play an important role in shaping tumor microenvironments; however, there is limited understanding of how natural killer (NK) cells can enhance radiation effects. This study aimed to assess the mechanism of reciprocal complementation of radiation and NK cells on tumor killing. Various tumor cell lines were co-cultured with human primary NK cells or NK cell line (NK-92) for short periods and then exposed to irradiation. Cell proliferation, apoptosis and transwell assays were performed to assess apoptotic efficacy and cell viability. Western blot analysis and immunoprecipitation methods were used to determine XIAP (X-linked inhibitor of apoptosis protein) and Smac (second mitochondria-derived activator of caspase) expression and interaction in tumor cells. Co-culture did not induce apoptosis in tumor cells, but a time- and dose-dependent enhancing effect was found when co-cultured cells were irradiated. A key role for caspase activation via perforin/granzyme B (Grz B) after cell-cell contact was determined, as the primary radiation enhancing effect. The efficacy of NK cell killing was attenuated by upregulation of XIAP to bind caspase-3 in tumor cells to escape apoptosis. Knockdown of XIAP effectively potentiated NK cell-mediated apoptosis. Radiation induced Smac released from mitochondria and neutralized XIAP and therefore increased the NK killing. Our findings suggest NK cells in tumor microenvironment have direct radiosensitization effect through Grz B injection while radiation enhances NK cytotoxicity through triggering Smac release.

Serum amyloid a as a predictive marker for radiation pneumonitis in lung cancer patients.

PURPOSE: To investigate serum markers associated with radiation pneumonitis (RP) grade ≥3 in patients with lung cancer who were treated with radiation therapy. METHODS AND MATERIALS: Pretreatment serum samples from patients with stage Ib-IV lung cancer who developed RP within 1 year after radiation therapy were analyzed to identify a proteome marker able to stratify patients prone to develop severe RP by surface-enhanced laser desorption/ionization time-of-flight mass spectrometry (SELDI-TOF-MS). Dosimetric parameters and 3 biological factors were compared. RESULTS: Serum samples from 16 patients (28%) with severe RP (grade 3-4) and 42 patients (72%) with no or mild RP (grade 0-2) were collected for analysis. All patients received a median of 54 Gy (range, 42-70 Gy) of three-dimensional conformal radiation therapy with a mean lung dose (MLD) of 1502 cGy (range, 700-2794 cGy). An m/z peak of 11,480 Da was identified by SELDI-TOF-MS, and serum amyloid A (SAA) was the primary splitter serum marker. The receiver operating characteristic area under the curve of SAA (0.94; 95% confidence interval [CI], 0.87-1.00) was higher than those of C-reactive protein (0.83; 95% CI, 0.72-0.94), interleukin-6 (0.79; 95% CI, 0.65-0.94), and MLD (0.57; 95% CI, 0.37-0.77). The best sensitivity and specificity of combined SAA and MLD for predicting RP were 88.9% and 96.0%, respectively. CONCLUSIONS: Baseline SAA could be used as an auxiliary marker for predicting severe RP. Extreme care should be taken to limit the lung irradiation dose in patients with high SAA.

Recombinant heat shock protein 70 in combination with radiotherapy as a source of tumor antigens to improve dendritic cell immunotherapy.

Local radiotherapy (RT) plus intratumoral dendritic cell (DC) injection can mediate immunological response. We hypothesized that co-injection of exogenous recombinant heat shock protein 70 (rHsp70) in combination with RT-DC could be as effective as co-injection of HSP-peptide for evoking specific immune response. rHsp70-prostate-specific antigen (rHSP70C'-PSA) and α-fetoprotein (rHSP70C'-AFP) were used to compare specific response. Growth inhibition of the tumor and the systemic anti-tumor immune response were measured on CT26/PSA and CT26/AFP mice model. Intratumoral co-injection of rHsp70 and DC into the irradiated tumor site induced a more potent anti-tumor immune response than injection of DC alone. rHsp70 was as effective as rHsp70C'-PSA or rHsp70C'-AFP in inducing a tumor-specific cytotoxic T lymphocyte response or tumor growth delay. These results demonstrate that co-administration with rHsp70 and RT could be a simple and effective source of tumor antigens to achieve RT-DC immunotherapy protocol and easy to apply in clinical use.

Improving Detection Accuracy of Lung Cancer Serum Proteomic Profiling via Two-Stage Training Process.

BACKGROUND: Surface-Enhanced Laser Desorption/Ionization Time-of-Flight Mass Spectrometry (SELDI-TOF-MS) is a frequently used technique for cancer biomarker research. The specificity of biomarkers detected by SELDI can be influenced by concomitant inflammation. This study aimed to increase detection accuracy using a two-stage analysis process. METHODS: Sera from 118 lung cancer patients, 72 healthy individuals, and 31 patients with inflammatory disease were randomly divided into training and testing groups by 3:2 ratio. In the training group, the traditional method of using SELDI profile analysis to directly distinguish lung cancer patients from sera was used. The two-stage analysis of distinguishing the healthy people and non-healthy patients (1st-stage) and then differentiating cancer patients from inflammatory disease patients (2nd-stage) to minimize the influence of inflammation was validated in the test group. RESULTS: In the test group, the one-stage method had 87.2% sensitivity, 37.5% specificity, and 64.4% accuracy. The two-stage method had lower sensitivity (> 70.1%) but statistically higher specificity (80%) and accuracy (74.7%). The predominantly expressed protein peak at 11480 Da was the primary splitter regardless of one- or two-stage analysis. This peak was suspected to be SAA (Serum Amyloid A) due to the similar m/z countered around this area. This hypothesis was further tested using an SAA ELISA assay. CONCLUSIONS: Inflammatory disease can severely interfere with the detection accuracy of SELDI profiles for lung cancer. Using a two-stage training process will improve the specificity and accuracy of detecting lung cancer.