Search

Deep learning (DL) can accelerate the prediction of prognostic biomarkers from routine pathology slides in colorectal cancer (CRC). However, current approaches rely on convolutional neural networks (CNNs) and have mostly been validated on small patient cohorts. Here, we develop a new transformer-based pipeline for end-to-end biomarker prediction from pathology slides by combining a pre-trained transformer encoder with a transformer network for patch aggregation. Our transformer-based approach substantially improves the performance, generalizability, data efficiency, and interpretability as compared with current state-of-the-art algorithms. After training and evaluating on a large multicenter cohort of over 13,000 patients from 16 colorectal cancer cohorts, we achieve a sensitivity of 0.99 with a negative predictive value of over 0.99 for prediction of microsatellite instability (MSI) on surgical resection specimens. We demonstrate that resection specimen-only training reaches clinical-grade performance on endoscopic biopsy tissue, solving a long-standing diagnostic problem.

PurposeAdvanced-stage mucinous ovarian carcinoma (MOC) has poor chemotherapy response and prognosis and lacks biomarkers to aid stage I adjuvant treatment. Differentiating primary MOC from gastrointestinal (GI) metastases to the ovary is also challenging due to phenotypic similarities. Clinicopathologic and gene-expression data were analyzed to identify prognostic and diagnostic features.Experimental designDiscovery analyses selected 19 genes with prognostic/diagnostic potential. Validation was performed through the Ovarian Tumor Tissue Analysis consortium and GI cancer biobanks comprising 604 patients with MOC (n = 333), mucinous borderline ovarian tumors (MBOT, n = 151), and upper GI (n = 65) and lower GI tumors (n = 55).ResultsInfiltrative pattern of invasion was associated with decreased overall survival (OS) within 2 years from diagnosis, compared with expansile pattern in stage I MOC [hazard ratio (HR), 2.77; 95% confidence interval (CI), 1.04-7.41, P = 0.042]. Increased expression of THBS2 and TAGLN was associated with shorter OS in MOC patients (HR, 1.25; 95% CI, 1.04-1.51, P = 0.016) and (HR, 1.21; 95% CI, 1.01-1.45, P = 0.043), respectively. ERBB2 (HER2) amplification or high mRNA expression was evident in 64 of 243 (26%) of MOCs, but only 8 of 243 (3%) were also infiltrative (4/39, 10%) or stage III/IV (4/31, 13%).ConclusionsAn infiltrative growth pattern infers poor prognosis within 2 years from diagnosis and may help select stage I patients for adjuvant therapy. High expression of THBS2 and TAGLN in MOC confers an adverse prognosis and is upregulated in the infiltrative subtype, which warrants further investigation. Anti-HER2 therapy should be investigated in a subset of patients. MOC samples clustered with upper GI, yet markers to differentiate these entities remain elusive, suggesting similar underlying biology and shared treatment strategies.

Purpose: Novel particle engineering approach was used in this study to generate high dose inhalable effervescent particles with synergistic effects against Pseudomonas aeruginosa biofilms. Methods: Spray dried co-amorphous salt of ciprofloxacin (CFX) and tartaric acid (TA) was prepared and coated with external layer of sodium bicarbonate and silica coated silver nanobeads. Design of experiments (DOE) was used to optimize physicochemical properties of particles for enhanced lung deposition. Results: Generated particles were co-amorphous CFX/TA showing that CFX lost its zwitterionic form and exhibiting distinct properties to CFX/HCl as assessed by FTIR and thermal analysis. Particles exhibited mass mean aerodynamic diameter (MMAD) of 3.3 μm, emitted dose of 78% and fine particle dose of 85%. Particles were further evaluated via antimicrobial assessment of minimum inhibitory concentrations (MIC) and minimum biofilm eradication concentration (MBEC). MIC and MBEC results showed that the hybrid particles were around 3–5 times more effective when compared to CFX signifying that synergistic effect was achieved. Diffusing wave spectroscopy results showed that the silver containing particles had a disruptive effect on rheological properties as opposed to silver free particles. Conclusions: Overall, these results showed the potential to use particle engineering to generate particles that are highly disruptive of bacterial biofilms.

Purpose: Novel particle engineering approach was used in this study to generate high dose inhalable effervescent particles with synergistic effects against Pseudomonas aeruginosa biofilms. Methods: Spray dried co-amorphous salt of ciprofloxacin (CFX) and tartaric acid (TA) was prepared and coated with external layer of sodium bicarbonate and silica coated silver nanobeads. Design of experiments (DOE) was used to optimize physicochemical properties of particles for enhanced lung deposition. Results: Generated particles were co-amorphous CFX/TA showing that CFX lost its zwitterionic form and exhibiting distinct properties to CFX/HCl as assessed by FTIR and thermal analysis. Particles exhibited mass mean aerodynamic diameter (MMAD) of 3.3 μm, emitted dose of 78% and fine particle dose of 85%. Particles were further evaluated via antimicrobial assessment of minimum inhibitory concentrations (MIC) and minimum biofilm eradication concentration (MBEC). MIC and MBEC results showed that the hybrid particles were around 3–5 times more effective when compared to CFX signifying that synergistic effect was achieved. Diffusing wave spectroscopy results showed that the silver containing particles had a disruptive effect on rheological properties as opposed to silver free particles. Conclusions: Overall, these results showed the potential to use particle engineering to generate particles that are highly disruptive of bacterial biofilms.

Purpose: Novel particle engineering approach was used in this study to generate high dose inhalable effervescent particles with synergistic effects against Pseudomonas aeruginosa biofilms. Methods: Spray dried co-amorphous salt of ciprofloxacin (CFX) and tartaric acid (TA) was prepared and coated with external layer of sodium bicarbonate and silica coated silver nanobeads. Design of experiments (DOE) was used to optimize physicochemical properties of particles for enhanced lung deposition. Results: Generated particles were co-amorphous CFX/TA showing that CFX lost its zwitterionic form and exhibiting distinct properties to CFX/HCl as assessed by FTIR and thermal analysis. Particles exhibited mass mean aerodynamic diameter (MMAD) of 3.3 μm, emitted dose of 78% and fine particle dose of 85%. Particles were further evaluated via antimicrobial assessment of minimum inhibitory concentrations (MIC) and minimum biofilm eradication concentration (MBEC). MIC and MBEC results showed that the hybrid particles were around 3–5 times more effective when compared to CFX signifying that synergistic effect was achieved. Diffusing wave spectroscopy results showed that the silver containing particles had a disruptive effect on rheological properties as opposed to silver free particles. Conclusions: Overall, these results showed the potential to use particle engineering to generate particles that are highly disruptive of bacterial biofilms.

Traditional image-based survival prediction models rely on discriminative patch labeling which make those methods not scalable to extend to large datasets. Recent studies have shown Multiple Instance Learning (MIL) framework is useful for histopathological images when no annotations are available in classification task. Different to the current image-based survival models that limit to key patches or clusters derived from Whole Slide Images (WSIs), we propose Deep Attention Multiple Instance Survival Learning (DeepAttnMISL) by introducing both siamese MI-FCN and attention-based MIL pooling to efficiently learn imaging features from the WSI and then aggregate WSI-level information to patient-level. Attention-based aggregation is more flexible and adaptive than aggregation techniques in recent survival models. We evaluated our methods on two large cancer whole slide images datasets and our results suggest that the proposed approach is more effective and suitable for large datasets and has better interpretability in locating important patterns and features that contribute to accurate cancer survival predictions. The proposed framework can also be used to assess individual patient's risk and thus assisting in delivering personalized medicine. Codes are available at https://github.com/uta-smile/DeepAttnMISL_MEDIA., Comment: 22 pages, 13 figures, published in Medical Image Analysis 65, 101789

Aberrant transcriptional silencing of genes by promoter hypermethylation is an important mechanism of tumour suppressor gene (TSG) inactivation in cancer. DNA hypomethylating agents such as 5-Aza-2’deoxycytidine (5-Aza-dC) are approved for the treatment of myelodysplastic syndrome and show promise in the treatment of solid tumours. These therapies are thought to exert their anti-tumour effects by reactivating hypermethylated genes. However, resistance to these drugs inevitably develops and this is associated with the silencing of genes initially re-expressed by therapy, termed gene re-silencing. The molecular basis of this is unknown, however understanding how it occurs may aid in the development of strategies to prevent or overcome resistance. We hypothesised that other chromatin modifications may silence gene expression in the absence of methylation. To study gene re-silencing following 5-Aza-dC exposure, a model gene (MLH1) and cell line (the colorectal carcinoma cell line RKO) were identified and validated. Gene reactivation was dependent on DNMT1 depletion and hypomethylation, nucleosome eviction and the acquisition of active histone marks at the MLH1 promoter. However, close inspection of the early stages of gene re-silencing showed that loss of expression occurred despite persistent hypomethylation. Instead, gene re-silencing coincided with the reassembly of nucleosomes at the MLH1 transcription start site, which preceded DNA re-methylation. Unexpectedly, changes in non-CpG methylation (methylation at CpT, CpA and CpC dinucleotides) were also observed, which we find may be attributable to alterations in the expression of the DNMT3A, DNMT3B or DNMT3L genes. Finally, despite the persistence of the repressive histone mark H3K9me3 following 5-Aza-dC treatment we find that, in this model, inhibitors of H3K9 methyltransferases do not delay gene re-silencing. These findings show that gene re-silencing is not driven by DNA re-methylation but involves other mechanisms

Cancer is caused by genetic and epigenetic changes resulting in altered patterns of gene expression. This study explored how somatic epigenetic events may predispose to the development of cancer and how genetic variation can underlie these events.Firstly, the role of constitutional methylation of the BRCA1 and BRCA2 genes in breast cancer development was investigated in BRCA-mutation negative cases of familial and early-onset breast cancer. No BRCA2 methylation was detected. Low level BRCA1 methylation was seen in 5/266 (2%) familial cases, 4/20 (20%) early-onset and 5/252 (2%) controls. No underlying sequence variants were found in regulatory regions of BRCA1 in methylation-positive cases. One familial case had two affected cousins who showed methylation (1-13%) – they all shared the same BRCA1 haplotype, suggesting a genetic basis for the epimutation. In another familial case, methylation detected in buccal cells (5%) was elevated in a cerebral metastasis (17%), suggesting a causative role in tumourigenesis in that case. However, methylation was found at similar low levels soma-wide and in the primary breast carcinoma in an early-onset case.Secondly, the role of the minor (T) allele of the MGMT enhancer rs16906252 C>T SNP, which correlates with MGMT promoter methylation, was studied in colorectal cancer (CRC). In a case-control study, the T allele was associated with increased risk of CRC showing MGMT-methylation, but not of developing CRC. Monoallelic methylation, mostly of the T allele, was observed in MGMT-methylated tumours retaining protein expression, producing discordant methylation and protein expression profiles. In peripheral blood and normal colorectal mucosa (NCM) of CRC cases and controls, low-level MGMT methylation was linked to the T allele. Allele quantification in NCM transcripts from cases and controls revealed a two-fold reduction in expression of the T compared to the C allele, possibly predisposing the T allele to methylation via altered inter

The use of a new sample holder for dynamic mechanical analysis (DMA) as a means to characterise the Tg of powdered hydroxypropyl methyl cellulose (HPMC) has been investigated. A sample holder was constructed consisting of a rectangular stainless steel container and a lid engineered to fit exactly within the walls of the container when clamped within a TA instruments Q800 DMA in dual cantilever configuration. Physical mixtures of HPMC (E4M) and aluminium oxide powders were placed in the holder and subjected to oscillating strains (1 Hz, 10 Hz and 100 Hz) whilst heated at 3 degrees C/min. The storage and loss modulus signals showed a large reduction in the mechanical strength above 150 degrees C which was attributed to a glass transition. Optimal experimental parameters were determined using a design of experiment procedure and by analysing the frequency dependence of Tg in Arrhenius plots. The parameters were a clamping pressure of 62 kPa, a mass ratio of 0.2 HPMC in aluminium oxide, and a loading mass of either 120 mg or 180 mg. At 1 Hz, a Tg of 177+/-1.2 degrees C (n=6) for powdered HPMC was obtained. In conclusion, the new powder holder was capable of measuring the Tg of pharmaceutical powders and a simple optimization protocol was established, useful in further applications of the DMA powder holder.

A modified AMS 28 test was given to advanced and upper-intermediate level ESOL students studying English Language at a Further Education College in the UK. The test measures student motivation and categorises it into six types plus amotivation. It is thought that identifying the type of motivation of the students may enable those with weak or no motivation to be identified and, if necessary, intervention to take place. It was shown that there was little significant difference in the background of the students who withdrew from the course compared with those who completed, although those students who had been resident in the country for 3-4 years were more likely to withdraw than those resident for either fewer or more years. Students in the cohort were found to be well motivated, and in general had an even balance of extrinsic and intrinsic motivational types. Students who withdrew were, however, more likely to have an extrinsic integrated type of motivation, which may indicate they placed value on the course as a means to getting (or getting better) employment in the UK, and left the course when this goal was achieved. No students in this cohort were amotivated, according to the test. The low number of students (13.5%) who withdrew from the course is an indication that students found the course of value and is similar to the national average for such courses, although this low number made statistical interpretation of the results difficult. It would be useful to extend this survey in the future to include students of lower levels who may become demotivated more quickly than the students in the higher levels questioned. It would also be of value to incorporate iii achievement statistics (ie which students obtained or did not obtain a full qualification) at the end of the course and correlate that data with motivation type.