Classifying patient and professional voice in social media health posts.

BACKGROUND: Patient-based analysis of social media is a growing research field with the aim of delivering precision medicine but it requires accurate classification of posts relating to patients' experiences. We motivate the need for this type of classification as a pre-processing step for further analysis of social media data in the context of related work in this area. In this paper we present experiments for a three-way document classification by patient voice, professional voice or other. We present results for a convolutional neural network classifier trained on English data from two different data sources (Reddit and Twitter) and two domains (cardiovascular and skin diseases). RESULTS: We found that document classification by patient voice, professional voice or other can be done consistently manually (0.92 accuracy). Annotators agreed roughly equally for each domain (cardiovascular and skin) but they agreed more when annotating Reddit posts compared to Twitter posts. Best classification performance was obtained when training two separate classifiers for each data source, one for Reddit and one for Twitter posts, when evaluating on in-source test data for both test sets combined with an overall accuracy of 0.95 (and macro-average F1 of 0.92) and an F1-score of 0.95 for patient voice only. CONCLUSION: The main conclusion resulting from this work is that combining social media data from platforms with different characteristics for training a patient and professional voice classifier does not result in best possible performance. We showed that it is best to train separate models per data source (Reddit and Twitter) instead of a model using the combined training data from both sources. We also found that it is preferable to train separate models per domain (cardiovascular and skin) while showing that the difference to the combined model is only minor (0.01 accuracy). Our highest overall F1-score (0.95) obtained for classifying posts as patient voice is a very good starting point for further analysis of social media data reflecting the experience of patients.

The leukocyte nuclear envelope proteome varies with cell activation and contains novel transmembrane proteins that affect genome architecture.

A favored hypothesis to explain the pathology underlying nuclear envelopathies is that mutations in nuclear envelope proteins alter genome/chromatin organization and thus gene expression. To identify nuclear envelope proteins that play roles in genome organization, we analyzed nuclear envelopes from resting and phytohemagglutinin-activated leukocytes because leukocytes have a particularly high density of peripheral chromatin that undergoes significant reorganization upon such activation. Thus, nuclear envelopes were isolated from leukocytes in the two states and analyzed by multidimensional protein identification technology using an approach that used expected contaminating membranes as subtractive fractions. A total of 3351 proteins were identified between both nuclear envelope data sets among which were 87 putative nuclear envelope transmembrane proteins (NETs) that were not identified in a previous proteomics analysis of liver nuclear envelopes. Nuclear envelope localization was confirmed for 11 new NETs using tagged fusion proteins and antibodies on spleen cryosections. 27% of the new proteins identified were unique to one or the other of the two leukocyte states. Differences in expression between activated and resting leukocytes were confirmed for some NETs by RT-PCR, and most of these proteins appear to only be expressed in certain types of blood cells. Several known proteins identified in both data sets have functions in chromatin organization and gene regulation. To test whether the novel NETs identified might include those that also regulate chromatin, nine were run through two screens for different chromatin effects. One screen found two NETs that can recruit a specific gene locus to the nuclear periphery, and the second found a different NET that promotes chromatin condensation. The variation in the protein milieu with pharmacological activation of the same cell population and consequences for gene regulation suggest that the nuclear envelope is a complex regulatory system with significant influences on genome organization.

Use of sequential chemical extractions to purify nuclear membrane proteins for proteomics identification.

The nuclear envelope (NE) is a double membrane system that is both a part of the endoplasmic reticulum and part of the nucleus. As its constituent proteins tend to be highly complexed with nuclear and cytoplasmic components, it is notoriously difficult to purify. Two methods can reduce this difficulty for the identification of nuclear membrane proteins: comparison to contaminating membranes and chemical extractions to enrich for certain groups of proteins. The purification of nuclear envelopes and contaminating microsomal membranes is described here along with procedures for chemical extraction using salt and detergent, chaotropes, or alkaline solutions. Each extraction method enriches for different combinations of nuclear envelope proteins. Finally, we describe the analysis of these fractions with MudPIT, a proteomics methodology that avoids gel extraction of bands to facilitate identification of minor proteins and membrane proteins that do not resolve well on gels. Together these three approaches can significantly increase the output of proteomics studies aimed at identifying the protein complement of subcellular membrane systems.

Human CLASP1 is an outer kinetochore component that regulates spindle microtubule dynamics.

One of the most intriguing aspects of mitosis is the ability of kinetochores to hold onto plus ends of microtubules that are actively gaining or losing tubulin subunits. Here, we show that CLASP1, a microtubule-associated protein, localizes preferentially near the plus ends of growing spindle microtubules and is also a component of a kinetochore region that we term the outer corona. A truncated form of CLASP1 lacking the kinetochore binding domain behaves as a dominant negative, leading to the formation of radial arrays of microtubule bundles that are highly resistant to depolymerization. Microinjection of CLASP1-specific antibodies suppresses microtubule dynamics at kinetochores and throughout the spindle, resulting in the formation of monopolar asters with chromosomes buried in the interior. Incubation with microtubule-stabilizing drugs rescues the kinetochore association with microtubule plus ends at the periphery of the asters. Our data suggest that CLASP1 is required at kinetochores for attached microtubules to exhibit normal dynamic behavior.

The cell cycle dependent mislocalisation of emerin may contribute to the Emery-Dreifuss muscular dystrophy phenotype.

Emerin is the nuclear membrane protein defective in X-linked Emery-Dreifuss muscular dystrophy (X-EDMD). The majority of X-EDMD patients have no detectable emerin. However, there are cases that produce mutant forms of emerin, which can be used to study its function. Our previous studies have shown that the emerin mutants S54F, P183T, P183H, Del95-99, Del236-241 (identified in X-EDMD patients) are targeted to the nuclear membrane but to a lesser extent than wild-type emerin. In this paper, we have studied how the mislocalisation of these mutant emerins may affect nuclear functions associated with the cell cycle using flow cytometry and immunofluorescence microscopy. We have established that cells expressing the emerin mutant Del236-241 (a deletion in the transmembrane domain), which was mainly localised in the cytoplasm, exhibited an aberrant cell cycle length. Thereafter, by examining the intracellular localisation of endogenously expressed lamin A/C and exogenously expressed wild-type and mutant forms of emerin after a number of cell divisions, we determined that the mutant forms of emerin redistributed endogenous lamin A/C. The extent of lamin A/C redistribution correlated with the amount of EGFP-emerin that was mislocalised. The amount of EGFP-emerin mislocalized, in turn, was associated with alterations in the nuclear envelope morphology. The nuclear morphology and redistribution of lamin A/C was most severely affected in the cells expressing the emerin mutant Del236-241. It is believed that emerin is part of a novel nuclear protein complex consisting of the barrier-to-autointegration factor (BAF), the nuclear lamina, nuclear actin and other associated proteins. The data presented here show that lamin A/C localisation is dominantly directed by its interaction with certain emerin mutants and perhaps wild-type emerin as well. These results suggest that emerin links A-type lamins to the nuclear envelope and that the correct localisation of these nuclear proteins is important for maintaining cell cycle timing.