Search

MotivationLarge Language Models (LLMs) are being adopted at an unprecedented rate, yet still face challenges in knowledge-intensive domains like biomedicine. Solutions such as pre-training and domain-specific fine-tuning add substantial computational overhead, requiring further domain-expertise. Here, we introduce a token-optimized and robust Knowledge Graph-based Retrieval Augmented Generation (KG-RAG) framework by leveraging a massive biomedical KG (SPOKE) with LLMs such as Llama-2-13b, GPT-3.5-Turbo and GPT-4, to generate meaningful biomedical text rooted in established knowledge.ResultsCompared to the existing RAG technique for Knowledge Graphs, the proposed method utilizes minimal graph schema for context extraction and uses embedding methods for context pruning. This optimization in context extraction results in more than 50% reduction in token consumption without compromising the accuracy, making a cost-effective and robust RAG implementation on proprietary LLMs. KG-RAG consistently enhanced the performance of LLMs across diverse biomedical prompts by generating responses rooted in established knowledge, accompanied by accurate provenance and statistical evidence (if available) to substantiate the claims. Further benchmarking on human curated datasets, such as biomedical true/false and multiple-choice questions (MCQ), showed a remarkable 71% boost in the performance of the Llama-2 model on the challenging MCQ dataset, demonstrating the framework's capacity to empower open-source models with fewer parameters for domain-specific questions. Furthermore, KG-RAG enhanced the performance of proprietary GPT models, such as GPT-3.5 and GPT-4. In summary, the proposed framework combines explicit and implicit knowledge of KG and LLM in a token optimized fashion, thus enhancing the adaptability of general-purpose LLMs to tackle domain-specific questions in a cost-effective fashion.Availability and implementationSPOKE KG can be accessed at https://spoke.rbvi.ucsf.edu/neighborhood.html. It can also be accessed using REST-API (https://spoke.rbvi.ucsf.edu/swagger/). KG-RAG code is made available at https://github.com/BaranziniLab/KG_RAG. Biomedical benchmark datasets used in this study are made available to the research community in the same GitHub repository.Supplementary informationSupplementary data are available at Bioinformatics online.

Knowledge representation and reasoning (KR&R) has been successfully implemented in many fields to enable computers to solve complex problems with AI methods. However, its application to biomedicine has been lagging in part due to the daunting complexity of molecular and cellular pathways that govern human physiology and pathology. In this article, we describe concrete uses of Scalable PrecisiOn Medicine Knowledge Engine (SPOKE), an open knowledge network that connects curated information from thirty-seven specialized and human-curated databases into a single property graph, with 3 million nodes and 15 million edges to date. Applications discussed in this article include drug discovery, COVID-19 research and chronic disease diagnosis, and management.

ObjectivesClinical notes are a veritable treasure trove of information on a patient's disease progression, medical history, and treatment plans, yet are locked in secured databases accessible for research only after extensive ethics review. Removing personally identifying and protected health information (PII/PHI) from the records can reduce the need for additional Institutional Review Boards (IRB) reviews. In this project, our goals were to: (1) develop a robust and scalable clinical text de-identification pipeline that is compliant with the Health Insurance Portability and Accountability Act (HIPAA) Privacy Rule for de-identification standards and (2) share routinely updated de-identified clinical notes with researchers.Materials and methodsBuilding on our open-source de-identification software called Philter, we added features to: (1) make the algorithm and the de-identified data HIPAA compliant, which also implies type 2 error-free redaction, as certified via external audit; (2) reduce over-redaction errors; and (3) normalize and shift date PHI. We also established a streamlined de-identification pipeline using MongoDB to automatically extract clinical notes and provide truly de-identified notes to researchers with periodic monthly refreshes at our institution.ResultsTo the best of our knowledge, the Philter V1.0 pipeline is currently the first and only certified, de-identified redaction pipeline that makes clinical notes available to researchers for nonhuman subjects' research, without further IRB approval needed. To date, we have made over 130 million certified de-identified clinical notes available to over 600 UCSF researchers. These notes were collected over the past 40 years, and represent data from 2757016 UCSF patients.

Kidney transplant recipients carrying the CYP3A5*1 allele have lower tacrolimus troughs, and higher dose requirements compared to those with the CYP3A5*3/*3 genotype. However, data on the effect of CYP3A5 alleles on post-transplant tacrolimus management are lacking. The effect of CYP3A5 metabolism phenotypes on the number of tacrolimus dose adjustments and troughs in the first 6 months post-transplant was evaluated in 78 recipients (64% Caucasians). Time to first therapeutic concentration, percentage of time in therapeutic range (TTR), and estimated glomerular filtration rate (eGFR) were also evaluated. Fifty-five kidney transplant recipients were CYP3A5 poor metabolizers (PM), 17 were intermediate metabolizers (IM), and 6 were extensive metabolizers (EM). Compared to PMs, EMs/IMs had significantly more dose adjustments (6.1 vs. 8.1, p = .015). Overall, 33.82% of trough measurements resulted in a dose change. There was no difference in the number of tacrolimus trough measurements between PMs and EM/IMs. The total daily tacrolimus dose requirements were higher in EMs and IMs compared to PMs (

A small percentage of bladder cancers in the general population have been found to harbor DNA viruses. In contrast, up to 25% of tumors of solid organ transplant recipients, who are at an increased risk of developing bladder cancer and have an overall poorer outcomes, harbor BK polyomavirus (BKPyV). To better understand the biology of the tumors and the mechanisms of carcinogenesis from potential oncoviruses, we performed whole genome and transcriptome sequencing on bladder cancer specimens from 43 transplant patients. Nearly half of the tumors from this patient population contained viral sequences. The most common were from BKPyV (N=9, 21%), JC polyomavirus (N=7, 16%), carcinogenic human papillomaviruses (N=3, 7%), and torque teno viruses (N=5, 12%). Immunohistochemistry revealed variable Large T antigen expression in BKPyV-positive tumors ranging from 100% positive staining of tumor tissue to less than 1%. In most cases of BKPyV-positive tumors, the viral genome appeared to be clonally integrated into the host chromosome consistent with microhomology-mediated end joining and coincided with focal amplifications of the tumor genome similar to other virus-mediated cancers. Significant changes in host gene expression consistent with the functions of BKPyV Large T antigen were also observed in these tumors. Lastly, we identified four mutation signatures in our cases, with those attributable to APOBEC3 and SBS5 being the most abundant. Mutation signatures associated with an antiviral drug, ganciclovir, and aristolochic acid, a nephrotoxic compound found in some herbal medicines, were also observed. The results suggest multiple pathways to carcinogenesis in solid organ transplant recipients with a large fraction being virus-associated.

MotivationKnowledge graphs (KGs) are being adopted in industry, commerce and academia. Biomedical KG presents a challenge due to the complexity, size and heterogeneity of the underlying information.ResultsIn this work, we present the Scalable Precision Medicine Open Knowledge Engine (SPOKE), a biomedical KG connecting millions of concepts via semantically meaningful relationships. SPOKE contains 27 million nodes of 21 different types and 53 million edges of 55 types downloaded from 41 databases. The graph is built on the framework of 11 ontologies that maintain its structure, enable mappings and facilitate navigation. SPOKE is built weekly by python scripts which download each resource, check for integrity and completeness, and then create a 'parent table' of nodes and edges. Graph queries are translated by a REST API and users can submit searches directly via an API or a graphical user interface. Conclusions/Significance: SPOKE enables the integration of seemingly disparate information to support precision medicine efforts.Availability and implementationThe SPOKE neighborhood explorer is available at https://spoke.rbvi.ucsf.edu.Supplementary informationSupplementary data are available at Bioinformatics online.

This chapter updates the COVID-19 chapter from the 2020 Annual Data Report with trends through February 12, 2022, and introduces trends in COVID-19-specific cause of death on the waiting list and posttransplant. Transplant rates remain at or above prepandemic levels for all organs, indicating a sustained transplantation system recovery following the initial 3-month disruption due to the onset of the pandemic. Posttransplant mortality and graft failure remain a concern in all organs, with rates surging corresponding to waves of the pandemic. Waitlist mortality due to COVID-19 is also a concern, particularly among kidney candidates. While the recovery of the transplantation system has been sustained in the second year of the pandemic, ongoing efforts should focus on reducing posttransplant and waitlist mortality due to COVID-19, and graft failure.

BACKGROUND: Focal segmental glomerulosclerosis (FSGS) is a common cause of end-stage kidney disease requiring kidney transplantation and can recur in the allograft in 30-80% of recipients resulting in reduced graft survival. Plasmapheresis has shown efficacy in treating some cases of recurrent FSGS but isolated plasmapheresis has not demonstrated efficacy in preventing recurrent FSGS. Rituximab has had anecdotal success in preventing recurrence in a single center study but has not been studied in combination with plasmapheresis for preventing FSGS recurrence. METHODS: We are conducting a randomized, controlled, multicenter clinical trial of adult and pediatric kidney transplant recipients with primary FSGS to assess whether plasmapheresis in combination with rituximab prevents recurrent disease post-transplantation. DISCUSSION: Rituximab combined with plasmapheresis is a promising, novel therapy to prevent recurrent FSGS, a disease with limited therapeutic options and no consensus guidelines for prevention or treatment. CLINICAL TRIAL REGISTRATION: https://clinicaltrials.gov/ct2/show/NCT03763643, identifier NCT03763643.

OBJECTIVE: To identify communication practices that clinicians can use to address racism faced by Black patients, build trusting relationships, and empower Black individuals in clinical care. DATA SOURCES: Qualitative data (N = 112 participants, August 2020-March 2021) collected in partnership with clinics primarily serving Black patients in Leeds, AL; Memphis, TN; Oakland, CA; and Rochester, NY. STUDY DESIGN: This multi-phased project was informed by human-centered design thinking and community-based participatory research principles. We mapped emergent communication and trust-building strategies to domains from the Presence 5 framework for fostering meaningful connection in clinical care. DATA COLLECTION METHODS: Interviews and focus group discussions explored anti-racist communication and patient-clinician trust (n = 36 Black patients; n = 40 nonmedical professionals; and n = 24 clinicians of various races and ethnicities). The Presence 5 Virtual National Community Advisory Board guided analysis interpretation. PRINCIPAL FINDINGS: The emergent Presence 5 for Racial Justice (P5RJ) practices include: (1) Prepare with intention by reflecting on identity, bias, and power dynamics; and creating structures to address bias and structural determinants of health; (2) Listen intently and completely without interruption and listen deeply for the potential impact of anti-Black racism on patient health and interactions with health care; (3) Agree on what matters most by having explicit conversations about patient goals, treatment comfort and consent, and referral planning; (4) Connect with the patients story, acknowledging socioeconomic factors influencing patient health and focusing on positive efforts; (5) Explore emotional cues by noticing and naming patient emotions, and considering how experiences with racism might influence emotions. CONCLUSION: P5RJ provides a framework with actionable communication practices to address pervasive racism experienced by Black patients. Effective implementation necessitates clinician self-reflection, personal commitment, and institutional support that offers time and resources to elicit a patients story and to address patient needs.

Potential regional variations in effects of COVID-19 on federally mandated, program-specific evaluations by the Scientific Registry of Transplant Recipients (SRTR) have been controversial. SRTR January 2022 program evaluations ended transplant follow-up on March 12, 2020, and excluded transplants performed from March 13, 2020 to June 12, 2020 (the "carve-out"). This study examined the carve-out's impact, and the effect of additionally censoring COVID-19 deaths, on first-year posttransplant outcomes for transplants from July 2018 through December 2020. Program-specific hazard ratios (HRs) for graft failure and death estimated under two alternative scenarios were compared with published HRs: (1) the carve-out was removed; (2) the carve-out was retained, but deaths due to COVID-19 were additionally censored. The HRs estimated by censoring COVID-19 deaths were highly correlated with those estimated with the carve-out alone (r2  = .96). Removal of the carve-out resulted in greater variation in HRs while remaining highly correlated (r2  = .82); however, little geographic impact of the carve-out was observed. The carve-out increased average HR in the Northwest by 0.049; carve-out plus censoring reduced average HR in the Midwest by 0.009. Other regions of the country were not significantly affected. Thus, the current COVID-19 carve-out does not appear to impart substantial bias based on the region of the country.

This study examines the experiences of students enrolled in a text-messaging advising program in order to understand the conditions for impact in this rapidly proliferating intervention model. The program under study was a 15-state text-messaging college advising trial that attempted to increase the college enrollment outcomes of over 30,000 students who attended U.S. high schools with large percentages of low-income students. Data came from 3600 advisees who responded to text-message queries about their experiences in the texting program. The content of the queries was informed by focus group responses from 18 program participants. Results indicate that text-message college advising offered students a combination of information, assistance, nudges, and emotional support that many students perceived kept them on track for a successful college process. Text-message advisees who were disengaged or critical of the program mistakenly believed that the advising was fully automated, had sufficient support elsewhere, or were not receiving the texts because of incorrect contact information. Study findings suggest that text messaging is best suited to providing advising on financial aid and other topics requiring specific information and concrete tasks. Students' reports of their experience in the program provide potential explanations for the mixed results that are beginning to be reported for texting programs and suggest implications for more effective designs. The study is useful as one of the first investigations of student responses to virtual advising, but its modest response rates suggest the need for new approaches to collecting evaluation data from participants in text-message advising campaigns.

Cannabis sativa emerges as a noteworthy candidate for its medicinal potential, particularly in wound healing. This review article explores the efficacy of cannabis oil in reducing reactive oxygen species (ROS) during the healing of acute and chronic wounds, comparing it to the standard treatments. ROS, produced from various internal and external sources, play a crucial role in wound development by causing cell and tissue damage. Understanding the role of ROS on skin wounds is essential, as they act both as signaling molecules and contributors to oxidative damage. Cannabis oil, recognized for its antioxidant properties, may help mitigate oxidative damage by scavenging ROS and upregulating antioxidative mechanisms, potentially enhancing wound healing. This review emphasizes ongoing research and the future potential of cannabis oil in dermatological treatments, highlighted through clinical studies and patent updates. Despite its promising benefits, optimizing cannabis oil formulations for therapeutic applications remains a challenge, underscoring the need for further research to realize its medicinal capabilities in wounds.

Context: Anti-Black racism is firmly rooted in US healthcare, but many clinicians do not have the tools and language to question their biases and address racism in clinical practice, eg biased communication practices such as non-compliance in medical documentation. Objective: Presence 5 for Racial Justice (P5RJ) leverages the Presence 5 patient-provider communication framework to identify anti-racism communication practices that support trusting relationships between physicians and Black patients and empower Black individuals in clinical care. Study design: For this multi-phased community-based participatory research (CBPR) overseen by an advisory board of clinicians and patients at four community clinics, we conducted a literature review, interviews with Black patients, clinician small-group discussions, and design thinking interviews with non-medical professionals. We mapped emergent communication practices to Presence 5 domains to create P5RJ. Setting: Four primary care clinics primarily serving Black patients in Oakland CA; Rochester NY; Leeds AL; Memphis TN. Population Studied: Total 113 participants (40 non-medical interviews, 37 Black patients interviewed, 12 advisory board members, 24 clinicians in discussion); 30 reviewed articles. Outcomes: Strategies on how providers, through communication and connection in the clinical visit, can navigate and address structural, institutional, and personally mediated forms of racism faced by Black patients. Results: P5RJ practices included: 1) Prepare with intention by reflecting on identity, bias, and power dynamics; and creating structures to address emergent bias and social determinants of health; 2) Listen intently and completely by using focused interpersonal listening without interruption and deep listening for racism impacts; give patients time and space to tell their story; 3) Agree on what matters most by having explicit conversations about patient goals, treatment comfort, consent, and referral planning; 4) Connect with the patients story by acknowledging socio-political factors influencing patient health and focusing on positive efforts/events to encourage patient agency; 5) Explore emotional cues by noticing and naming patient emotions and considering how racial trauma might influence these emotions. Conclusion: P5RJ practices offer strategies to reflect on clinician biases, address racism and known gaps in care for Black patients and promote health equity in their clinical care.

Synthetic biology has established powerful tools to precisely control cell function. Engineering these systems to meet clinical requirements has enormous medical implications. Here, we adopted a clinically driven design process to build receptors for the autonomous control of therapeutic cells. We examined the function of key domains involved in regulated intramembrane proteolysis and showed that systematic modular engineering can generate a class of receptors that we call synthetic intramembrane proteolysis receptors (SNIPRs) that have tunable sensing and transcriptional response abilities. We demonstrate the therapeutic potential of the receptor platform by engineering human primary T cells for multi-antigen recognition and production of dosed, bioactive payloads relevant to the treatment of disease. Our design framework enables the development of fully humanized and customizable transcriptional receptors for the programming of therapeutic cells suitable for clinical translation.

IntroductionAnti-Black racism has strong roots in American health care and medical education. While curricula on social determinants of health are increasingly common in medical training, curricula directly addressing anti-Black racism are limited. Existing frameworks like the Presence 5 framework for humanism in medicine can be adapted to develop a novel workshop that promotes anti-racism communication.MethodsWe performed a literature review of anti-racism collections and categorized anti-racism communication practices using the Presence 5 framework to develop the Presence 5 for Racial Justice Workshop. Implementation included an introductory didactic, a small-group discussion, and a large-group debrief. Participants evaluated the workshop via an online survey, and we analyzed the resulting qualitative feedback.ResultsA total of 17 participants took part in two workshops, with nine of the participants responding to the evaluation survey. Themes that emerged from survey responses included strengths of and improvements for the workshop structure (protected time for anti-racism discussion, dialogue between learners and faculty) and content (specific phrases and language, practicing self-reflection).DiscussionThe workshop provides participants with a semistructured discussion around the five anti-racism communication practices. Barriers to implementation include incorporating the workshop into existing curricula and ensuring diverse learners. Barriers to evaluating the workshop include the low survey response rate. Recommendations to improve the workshop include using case-based discussion and varying the workshop structure according to institutional needs. Next steps include an implementation study to evaluate the acceptability, feasibility, and effectiveness of the workshop.

Knowledge representation and reasoning (KR&R) has been successfully implemented in many fields to enable computers to solve complex problems with AI methods. However, its application to biomedicine has been lagging in part due to the daunting complexity of molecular and cellular pathways that govern human physiology and pathology. In this article we describe concrete uses of SPOKE, an open knowledge network that connects curated information from 37 specialized and human-curated databases into a single property graph, with 3 million nodes and 15 million edges to date. Applications discussed in this article include drug discovery, COVID-19 research and chronic disease diagnosis and management.