RESUMO
Accurate identification of human leukocyte antigen (HLA) alleles is essential for various clinical and research applications, such as transplant matching and drug sensitivities. Recent advances in RNA-seq technology have made it possible to impute HLA types from sequencing data, spurring the development of a large number of computational HLA typing tools. However, the relative performance of these tools is unknown, limiting the ability for clinical and biomedical research to make informed choices regarding which tools to use. Here we report the study design of a comprehensive benchmarking of the performance of 12 HLA callers across 682 RNA-seq samples from 8 datasets with molecularly defined gold standard at 5 loci, HLA-A, -B, -C, -DRB1, and -DQB1. For each HLA typing tool, we will comprehensively assess their accuracy, compare default with optimized parameters, and examine for discrepancies in accuracy at the allele and loci levels. We will also evaluate the computational expense of each HLA caller measured in terms of CPU time and RAM. We also plan to evaluate the influence of read length over the HLA region on accuracy for each tool. Most notably, we will examine the performance of HLA callers across European and African groups, to determine discrepancies in accuracy associated with ancestry. We hypothesize that RNA-Seq HLA callers are capable of returning high-quality results, but the tools that offer a good balance between accuracy and computational expensiveness for all ancestry groups are yet to be developed. We believe that our study will provide clinicians and researchers with clear guidance to inform their selection of an appropriate HLA caller.
RESUMO
Synthetic development is a nascent field of research that uses the tools of synthetic biology to design genetic programs directing cellular patterning and morphogenesis in higher eukaryotic cells, such as mammalian cells. One specific example of such synthetic genetic programs was based on cell-cell contact-dependent signaling using synthetic Notch pathways and was shown to drive the formation of multilayered spheroids by modulating cell-cell adhesion via differential expression of cadherin family proteins in a mouse fibroblast cell line (L929). The design method for these genetic programs relied on trial and error, which limited the number of possible circuits and parameter ranges that could be explored. Here, we build a parameterized computational framework that, given a cell-cell communication network driving changes in cell adhesion and initial conditions as inputs, predicts developmental trajectories. We first built a general computational framework where contact-dependent cell-cell signaling networks and changes in cell-cell adhesion could be designed in a modular fashion. We then used a set of available in vitro results (that we call the "training set" in analogy to similar pipelines in the machine learning field) to parameterize the computational model with values for adhesion and signaling. We then show that this parameterized model can qualitatively predict experimental results from a "testing set" of available in vitro data that varied the genetic network in terms of adhesion combinations, initial number of cells, and even changes to the network architecture. Finally, this parameterized model is used to recommend novel network implementation for the formation of a four-layered structure that has not been reported previously. The framework that we develop here could function as a testing ground to identify the reachable space of morphologies that can be obtained by controlling contact-dependent cell-cell communications and adhesion with these molecular tools and in this cellular system. Additionally, we discuss how the model could be expanded to include other forms of communication or effectors for the computational design of the next generation of synthetic developmental trajectories.
