Optimising the use of electronic medical records for large scale research in psychiatry.

The explosion and abundance of digital data could facilitate large-scale research for psychiatry and mental health. Research using so-called "real world data"-such as electronic medical/health records-can be resource-efficient, facilitate rapid hypothesis generation and testing, complement existing evidence (e.g. from trials and evidence-synthesis) and may enable a route to translate evidence into clinically effective, outcomes-driven care for patient populations that may be under-represented. However, the interpretation and processing of real-world data sources is complex because the clinically important 'signal' is often contained in both structured and unstructured (narrative or "free-text") data. Techniques for extracting meaningful information (signal) from unstructured text exist and have advanced the re-use of routinely collected clinical data, but these techniques require cautious evaluation. In this paper, we survey the opportunities, risks and progress made in the use of electronic medical record (real-world) data for psychiatric research.

Clinical Prompt Learning With Frozen Language Models.

When the first transformer-based language models were published in the late 2010s, pretraining with general text and then fine-tuning the model on a task-specific dataset often achieved the state-of-the-art performance. However, more recent work suggests that for some tasks, directly prompting the pretrained model matches or surpasses fine-tuning in performance with few or no model parameter updates required. The use of prompts with language models for natural language processing (NLP) tasks is known as prompt learning. We investigated the viability of prompt learning on clinically meaningful decision tasks and directly compared this with more traditional fine-tuning methods. Results show that prompt learning methods were able to match or surpass the performance of traditional fine-tuning with up to 1000 times fewer trainable parameters, less training time, less training data, and lower computation resource requirements. We argue that these characteristics make prompt learning a very desirable alternative to traditional fine-tuning for clinical tasks, where the computational resources of public health providers are limited, and where data can often not be made available or not be used for fine-tuning due to patient privacy concerns. The complementary code to reproduce the experiments presented in this work can be found at https://github.com/NtaylorOX/Public_Clinical_Prompt.

A HaemAtlas: characterizing gene expression in differentiated human blood cells.

Hematopoiesis is a carefully controlled process that is regulated by complex networks of transcription factors that are, in part, controlled by signals resulting from ligand binding to cell-surface receptors. To further understand hematopoiesis, we have compared gene expression profiles of human erythroblasts, megakaryocytes, B cells, cytotoxic and helper T cells, natural killer cells, granulocytes, and monocytes using whole genome microarrays. A bioinformatics analysis of these data was performed focusing on transcription factors, immunoglobulin superfamily members, and lineage-specific transcripts. We observed that the numbers of lineage-specific genes varies by 2 orders of magnitude, ranging from 5 for cytotoxic T cells to 878 for granulocytes. In addition, we have identified novel coexpression patterns for key transcription factors involved in hematopoiesis (eg, GATA3-GFI1 and GATA2-KLF1). This study represents the most comprehensive analysis of gene expression in hematopoietic cells to date and has identified genes that play key roles in lineage commitment and cell function. The data, which are freely accessible, will be invaluable for future studies on hematopoiesis and the role of specific genes and will also aid the understanding of the recent genome-wide association studies.

Association scan of 14,500 nonsynonymous SNPs in four diseases identifies autoimmunity variants.

We have genotyped 14,436 nonsynonymous SNPs (nsSNPs) and 897 major histocompatibility complex (MHC) tag SNPs from 1,000 independent cases of ankylosing spondylitis (AS), autoimmune thyroid disease (AITD), multiple sclerosis (MS) and breast cancer (BC). Comparing these data against a common control dataset derived from 1,500 randomly selected healthy British individuals, we report initial association and independent replication in a North American sample of two new loci related to ankylosing spondylitis, ARTS1 and IL23R, and confirmation of the previously reported association of AITD with TSHR and FCRL3. These findings, enabled in part by increased statistical power resulting from the expansion of the control reference group to include individuals from the other disease groups, highlight notable new possibilities for autoimmune regulation and suggest that IL23R may be a common susceptibility factor for the major 'seronegative' diseases.