CD8+ T-cell Differentiation and Dysfunction Inform Treatment Response in Acute Myeloid Leukemia

The interplay between T-cell states of differentiation, dysfunction, and treatment response in acute myeloid leukemia (AML) remains unclear. Here, we leveraged a multimodal approach encompassing high-dimensional flow cytometry and single-cell transcriptomics and found that early memory CD8+ T cells are associated with therapy response and exhibit a bifurcation into two distinct terminal end states. One state is enriched for markers of activation, whereas the other expresses NK-like and senescence markers. The skewed clonal differentiation trajectory towards CD8+ senescence was also a hallmark indicative of therapy resistance. We validated these findings by generating an AML CD8+ single-cell atlas integrating our data and other independent datasets. Finally, our analysis revealed that an imbalance between CD8+ early memory and senescent-like cells is linked to AML treatment refractoriness and poor survival. Our study provides crucial insights into the dynamics of CD8+ T-cell differentiation and advances our understanding of CD8+ T-cell dysfunction in AML.

The Unintended Consequences of Discount Regularization: Improving Regularization in Certainty Equivalence Reinforcement Learning

Discount regularization, using a shorter planning horizon when calculating the optimal policy, is a popular choice to restrict planning to a less complex set of policies when estimating an MDP from sparse or noisy data (Jiang et al., 2015). It is commonly understood that discount regularization functions by de-emphasizing or ignoring delayed effects. In this paper, we reveal an alternate view of discount regularization that exposes unintended consequences. We demonstrate that planning under a lower discount factor produces an identical optimal policy to planning using any prior on the transition matrix that has the same distribution for all states and actions. In fact, it functions like a prior with stronger regularization on state-action pairs with more transition data. This leads to poor performance when the transition matrix is estimated from data sets with uneven amounts of data across state-action pairs. Our equivalence theorem leads to an explicit formula to set regularization parameters locally for individual state-action pairs rather than globally. We demonstrate the failures of discount regularization and how we remedy them using our state-action-specific method across simple empirical examples as well as a medical cancer simulator.

Learning Predictive and Interpretable Timeseries Summaries from ICU Data

Machine learning models that utilize patient data across time (rather than just the most recent measurements) have increased performance for many risk stratification tasks in the intensive care unit. However, many of these models and their learned representations are complex and therefore difficult for clinicians to interpret, creating challenges for validation. Our work proposes a new procedure to learn summaries of clinical timeseries that are both predictive and easily understood by humans. Specifically, our summaries consist of simple and intuitive functions of clinical data (e.g. "falling mean arterial pressure"). Our learned summaries outperform traditional interpretable model classes and achieve performance comparable to state-of-the-art deep learning models on an in-hospital mortality classification task.

Optimizing for Interpretability in Deep Neural Networks with Tree Regularization

Deep models have advanced prediction in many domains, but their lack of interpretability  remains a key barrier to the adoption in many real world applications. There exists a large  body of work aiming to help humans understand these black box functions to varying levels  of granularity – for example, through distillation, gradients, or adversarial examples. These  methods however, all tackle interpretability as a separate process after training. In this  work, we take a different approach and explicitly regularize deep models so that they are  well-approximated by processes that humans can step through in little time. Specifically,  we train several families of deep neural networks to resemble compact, axis-aligned decision  trees without significant compromises in accuracy. The resulting axis-aligned decision  functions uniquely make tree regularized models easy for humans to interpret. Moreover,  for situations in which a single, global tree is a poor estimator, we introduce a regional tree regularizer that encourages the deep model to resemble a compact, axis-aligned decision  tree in predefined, human-interpretable contexts. Using intuitive toy examples, benchmark  image datasets, and medical tasks for patients in critical care and with HIV, we demonstrate  that this new family of tree regularizers yield models that are easier for humans to simulate  than L1 or L2 penalties without sacrificing predictive power. 

Preferential Mixture-of-Experts: Interpretable Models that Rely on Human Expertise As Much As Possible

We propose Preferential MoE, a novel human-ML mixture-of-experts model that augments human expertise in decision making with a data-based classifier only when necessary for predictive performance. Our model exhibits an interpretable gating function that provides information on when human rules should be followed or avoided. The gating function is maximized for using human-based rules, and classification errors are minimized. We propose solving a coupled multi-objective problem with convex subproblems. We develop approximate algorithms and study their performance and convergence. Finally, we demonstrate the utility of Preferential MoE on two clinical applications for the treatment of Human Immunodeficiency Virus (HIV) and management of Major Depressive Disorder (MDD).

Real-time prediction of COVID-19 related mortality using electronic health records

Coronavirus disease 2019 (COVID-19) is a respiratory disease with rapid human-to-human transmission caused by the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). Due to the exponential growth of infections, identifying patients with the highest mortality risk early is critical to enable effective intervention and prioritisation of care. Here, we present the COVID-19 early warning system (CovEWS), a risk scoring system for assessing COVID-19 related mortality risk that we developed using data amounting to a total of over 2863 years of observation time from a cohort of 66 430 patients seen at over 69 healthcare institutions. On an external cohort of 5005 patients, CovEWS predicts mortality from 78.8% (95% confidence interval [CI]: 76.0, 84.7%) to 69.4% (95% CI: 57.6, 75.2%) specificity at sensitivities greater than 95% between, respectively, 1 and 192 h prior to mortality events. CovEWS could enable earlier intervention, and may therefore help in preventing or mitigating COVID-19 related mortality.

Host Genomics of the HIV-1 Reservoir Size and Its Decay Rate During Suppressive Antiretroviral Treatment

BACKGROUND

The primary hurdle for the eradication of HIV-1 is the establishment of a latent viral reservoir early after primary infection. Here, we investigated the potential influence of human genetic variation on the HIV-1 reservoir size and its decay rate during suppressive antiretroviral treatment.

SETTING

Genome-wide association study and exome sequencing study to look for host genetic determinants of HIV-1 reservoir measurements in patients enrolled in the Swiss HIV Cohort Study, a nation-wide prospective observational study.

METHODS

We measured total HIV-1 DNA in peripheral blood mononuclear cells from study participants, as a proxy for the reservoir size at 3 time points over a median of 5.4 years, and searched for associations between human genetic variation and 2 phenotypic readouts: the reservoir size at the first time point and its decay rate over the study period. We assessed the contribution of common genetic variants using genome-wide genotyping data from 797 patients with European ancestry enrolled in the Swiss HIV Cohort Study and searched for a potential impact of rare variants and exonic copy number variants using exome sequencing data generated in a subset of 194 study participants.

RESULTS

Genome-wide and exome-wide analyses did not reveal any significant association with the size of the HIV-1 reservoir or its decay rate on suppressive antiretroviral treatment.

CONCLUSIONS

Our results point to a limited influence of human genetics on the size of the HIV-1 reservoir and its long-term dynamics in successfully treated individuals.

Information Bottleneck for Estimating Treatment Effects with Systematically Missing Covariates

Estimating the effects of an intervention from high-dimensional observational data is a challenging problem due to the existence of confounding. The task is often further complicated in healthcare applications where a set of observations may be entirely missing for certain patients at test time, thereby prohibiting accurate inference. In this paper, we address this issue using an approach based on the information bottleneck to reason about the effects of interventions. To this end, we first train an information bottleneck to perform a low-dimensional compression of covariates by explicitly considering the relevance of information for treatment effects. As a second step, we subsequently use the compressed covariates to perform a transfer of relevant information to cases where data are missing during testing. In doing so, we can reliably and accurately estimate treatment effects even in the absence of a full set of covariate information at test time. Our results on two causal inference benchmarks and a real application for treating sepsis show that our method achieves state-of-the-art performance, without compromising interpretability.

Determinants of HIV-1 reservoir size and long-term dynamics during suppressive ART

The HIV-1 reservoir is the major hurdle to a cure. We here evaluate viral and host characteristics associated with reservoir size and long-term dynamics in 1,057 individuals on suppressive antiretroviral therapy for a median of 5.4 years. At the population level, the reservoir decreases with diminishing differences over time, but increases in 26.6% of individuals. Viral blips and low-level viremia are significantly associated with slower reservoir decay. Initiation of ART within the first year of infection, pretreatment viral load, and ethnicity affect reservoir size, but less so long-term dynamics. Viral blips and low-level viremia are thus relevant for reservoir and cure studies.

Here, Bachmann et al. provide data on long-term dynamics of the HIV-1 reservoir in 1,057 individuals on suppressive antiretroviral therapy and show that in 26.6% of individuals the reservoir increases. Viral blips and low-level viremia are significantly associated with a slower reservoir decay.

Improving counterfactual reasoning with kernelised dynamic mixing models

Simulation-based approaches to disease progression allow us to make counterfactual predictions about the effects of an untried series of treatment choices. However, building accurate simulators of disease progression is challenging, limiting the utility of these approaches for real world treatment planning. In this work, we present a novel simulation-based reinforcement learning approach that mixes between models and kernel-based approaches to make its forward predictions. On two real world tasks, managing sepsis and treating HIV, we demonstrate that our approach both learns state-of-the-art treatment policies and can make accurate forward predictions about the effects of treatments on unseen patients.

Combining Kernel and Model Based Learning for HIV Therapy Selection

We present a mixture-of-experts approach for HIV therapy selection. The heterogeneity in patient data makes it difficult for one particular model to succeed at providing suitable therapy predictions for all patients. An appropriate means for addressing this heterogeneity is through combining kernel and model-based techniques. These methods capture different kinds of information: kernel-based methods are able to identify clusters of similar patients, and work well when modelling the viral response for these groups. In contrast, model-based methods capture the sequential process of decision making, and are able to find simpler, yet accurate patterns in response for patients outside these groups. We take advantage of this information by proposing a mixture-of-experts model that automatically selects between the methods in order to assign the most appropriate therapy choice to an individual. Overall, we verify that therapy combinations proposed using this approach significantly outperform previous methods.

Diagnostic accuracy of mammography and scintimammography in detection of primary breast cancer related size of the tumour

BACKGROUND

The detection of breast cancer relies on physical examination and mammography (XMM). The sensitivity of conventional imaging is lesions-size dependent. However, mammography has good sensitivity in small tumours when microcalcification is present, but the sensitivity is relatively low in other small tumours. The use of scintimammography with Tc-99m sestaMIBI (SMM) may be helpful in this situation. The aim of this study is to compare the diagnostic accuracy of mammography and scintimammography across all groups with different sizes of lesions.

METHODS

Data for 273 patients aged 26 to 84 years (mean 52; SD 12) with 298 lesions were analysed. All lesions were removed and final histology and pathological size of the tumours were retrospectively reviewed (mean size 2.9 cm, SD 2.4 cm). All patients had XMM and SMM. Each imaging study was graded using five grades of certainty. SMM images were reviewed by an independent specialist blinded to clinical presentation and XMM results. The diagnostic accuracy of each test, overall and by each lesions size subgroup, was compared using receiver operating characteristic curve (ROC) analysis.

RESULTS

The sensitivity of scintimammography was consistently high across all size groups; overall 89% (82?100%) compared with 68% (51?88%) for mammography. In those tumours below 2 cm, the area under the ROC was significantly larger with SMM than XMM (p < 0.05), with 80% certainty indicative of greater diagnostic accuracy. Also in all patients the area under the ROC was significantly larger with SMM than XMM (p < 0.05), with 90% certainty indicative of greater diagnostic accuracy. In other groups of lesions size the area under ROC for SMM was larger in lesions size between 2 and 4 cm and slightly smaller in lesions over 4 cm, but it was not significant.

CONCLUSION

Scintimammography using Tc-99m sestaMIBI seems to be a sensitive and consistently reliable diagnostic test for breast cancer independent of the size of the tumour.

Response of women aged 65-74 to invitation for screening for breast cancer by mammography: a pilot study in London, UK

OBJECTIVE

To investigate the response and benefits to be gained from mammographic screening for breast cancer in women aged 65-74, who are not normally invited for screening.

DESIGN

This was a pilot study comprising women aged 65-74 who are not currently invited for routine screening under the NHS breast screening programme. The results from this study were compared with the results of routinely screened women (aged 50-64) from the same health district.

SETTING

A mobile breast screening unit in the grounds of the Royal Free Hospital.

SUBJECTS

These comprised 5004 women aged 65-74 registered with GPs in the district of Hampstead and on the family health services authority (FHSA) lists. A total of 168 (3.4%) were initially excluded by the general practitioner or FHSA, and 286 (5.9%) of the invitation letters were returned by the Post Office or by other people as not deliverable for some reason.

MAIN OUTCOME MEASURES

Response rates to the invitation were assessed using three indices: crude population coverage rate, crude invited population coverage rate, and corrected invited population coverage rate.

RESULTS

With regard to response rates, 1684 women aged 65-74 (37% of all those invited, excluding those who were not available) were screened, compared with 2894 (42%) women aged 50-64. The three response rates were higher for younger women than older: the crude population coverage rate was 37.1%, the crude invited population coverage rate was 38.9%, and the corrected invited population coverage rate was 42.1% for women aged 50-64, compared with 32.9%, 34.4%, and 36.8% respectively for women aged 65-69 and 34.3%, 35.2%, and 37.2% for women aged 70-74. The rate of assessment increased significantly with increasing age, with 3.18% of the 50-64 population screened being assessed, as compared with 4.14% and 4.83% of the women aged 65-69 and 70-74, respectively. Most biopsies done in the older women gave positive results, as did the biopsies from the 50-64 age group. However, the biopsy rate increased significantly with increasing age. The cancer detection rates in the women aged 65-69 and 70-74 were 14.2/1000 and 13.2/1000 compared with an incident screening round rate of 4.5/1000 in women aged 50-64.

CONCLUSIONS

These results show that there is potential for similar attendance at routine screening by older women if they are invited in the same way as younger women. As the assessment, biopsy, and cancer detection rates in the older women are significantly higher than in the 50-64 year olds, the costs and benefits of including them in the NHS screening programme should be reassessed.

Family history and risk of breast cancer

The risk of breast cancer in first degree relatives of patients with breast cancer can be derived from family history and is dependent upon the age at diagnosis in the index patient. For the relatives of index patients older than 55, the relative risk is 1.57, if less than 55 the relative risk is 2.29, and 3.85 if less than 45 (95% confidence limits 0.83 to 2.68, 1.18 to 4.01, and 1.67 to 3.85, respectively). First degree relatives of patients with bilateral breast cancer have a 6.43-fold increase in risk (95% confidence limits 1.32 to 18.77). The genetic contribution to overall lifetime liability to breast cancer in the relatives declines rapidly with increasing age of onset of breast cancer in the index patient from 37% at 20 years to 8% by 45 years. This information can be used in clinical practice for counselling and the establishment of screening programmes.

Identification of bone and liver metastases from breast cancer by measurement of plasma alkaline phosphatase isoenzyme activity

Plasma alkaline phosphatase isoenzyme activities were determined in patients with breast cancer to diagnose and monitor bone and liver metastases. Bone alkaline phosphatase activity was increased in 21 of 50 patients (42%) with radiologically confirmed bone metastases, while total alkaline phosphatase activity was increased in only 10 of 50 (20%); liver alkaline phosphatase activity was raised in 12 of 25 patients (48%) with liver metastases. All patients with liver metastases had bone metastases. Bone alkaline phosphatase activity was significantly higher in patients with symptomatic bone disease. Isoenzyme determination provided additional information that would have changed patient management in five of 20 patients who were monitored serially. Measurement of alkaline phosphatase isoenzyme activity, though less sensitive than imaging procedures, can assist in screening for, and in early detection of, a high proportion of bone and liver metastases, and can provide useful objective evidence of their response to treatment.

Measurement of biliary alkaline phosphatase by mini-column chromatography and by electrophoresis and its application to the detection of liver metastases in patients with breast cancer

We describe a simple, rapid, and reproducible ion exchange mini-column chromatographic method for the quantitative measurement of biliary alkaline phosphatase in plasma. We have used this method to evaluate a cellulose acetate electrophoretic method, which was used to assess the value of measuring biliary alkaline phosphatase in 85 patients with breast cancer investigated for possible hepatic metastases. Biliary alkaline phosphatase activity was abnormal in 19 of 24 patients (79%) with liver metastases, but abnormalities were also found in 12 of 61 patients (20%) without hepatic metastases; in only 37% of patients with positive test results was this a consequence of liver metastases. For the identification of liver metastases, therefore, the method has useful sensitivity but limited specificity. Neither sensitivity nor specificity was significantly better than that of plasma gamma-glutamyltransferase activity, which was measured concomitantly.

The effect of porta-systemic shunting on liver, brain and kidney zinc concentrations in the rat

Patients with alcoholic cirrhosis have subnormal liver zinc concentrations, and excessive urinary zinc excretion. It has been suggested that diversion of dietary zinc into the systemic circulation through porta-systemic shunts may be a major factor influencing the organ distribution of zinc in these patients. To test this hypothesis, we randomized Sprague-Dawley rats into a sham-operated group, and a group subjected to partial portal-vein occlusion (PPVO), which induces the formation of extensive porta-systemic collaterals. Three months after the operation, liver, brain and kidney zinc concentrations were measured in 8 sham-operated and 8 PPVO rats. Mean liver zinc concentrations were significantly higher in the PPVO group when compared to sham-operated controls (P less than 0.05). Brain and kidney zinc concentrations were similar in the 2 groups. This short-term study indicates that liver zinc concentrations are maintained and even increased in rats with porta-systemic shunting, and that diversion of portal blood into the systemic circulation is not a major cause of the low liver zinc concentration observed in alcoholic cirrhosis.

The effect of oral copper loading and portasystemic shunting on the distribution of copper in the liver, brain, kidney, and cornea of the rat

In Wilson's disease, redistribution of copper from the liver to extrahepatic tissue coincides with the development of liver disease and cirrhosis. We have considered the possibility that portasystemic shunting may be a factor determining the organ distribution of copper in patients with liver disease. Thirty-two Sprague-Dawley rats were randomized into a sham-operated group, and a group subjected to partial occlusion of the portal vein (PPVO). Half the rats in each group were fed on normal diet and the remainder had copper added to the drinking water. Rats subjected to PPVO developed large portasystemic shunts. Over a 3-month period, none of the groups showed an increase in serum copper concentrations, or copper oxidase levels. Liver and kidney copper concentrations were similarly increased in copper supplemented sham-operated and PPVO rats, but brain copper content in both groups was similar to the unsupplemented controls. In neither unsupplemented nor copper supplemented rats subjected to PPVO, was the organ distribution of copper different from sham-operated controls. This study suggests that portasystemic shunting does not alter the organ distribution, or tissue concentration of copper.