🤔 Aba | 全自动biomarker分析神包！~（原作者用这个包发了三篇Nature啦~）

1写在前面

今天介绍一个Github上的神包吧, 主要是用于Biomarker的临床分析, 原作者用这个包已经发了3篇Nature了, 一起看看吧:👇

在一些针对Biomarker的临床研究中, 我们常常需对Biomarker进行模型拟合, 预测效果评估等等.🥰 这个包可以完美解决这些问题, 并且直接输出发表级图表, 简单介绍一下用法给大家吧.🥳

2用到的包

rm(list = ls())
#devtools::install_github("ncullen93/abaR")
library(aba)
library(tidyverse)
library(ggsci)

3示例数据

dat <- adnimerge %>% 
  dplyr::filter(VISCODE == 'bl')

DT::datatable(dat)

4变量一览

看一下都有什么变量, 都是什么类型. 🤞

str(dat)

5建立模型

5.1 原函数

大家可以通过这种方式来建立模型.

  aba_model(
   data = NULL,
   groups = NULL,
   outcomes = NULL,
   predictors = NULL,
   covariates = NULL,
   stats = NULL,
   evals = NULL,
   include_basic = TRUE
 )

5.2 pipeline形式

这里也提供了pipeline的形式来编写代码, 个人也是更倾向于这种方式, 大家试一下吧. 😏

我们在这里设定两个结局指标: ConvertedToAlzheimers和CSF_ABETA_STATUS_bl.🤒 两个预测指标, 即Biomarker, PLASMA_PTAU181_bl和PLASMA_NFL_bl. 🤫

模型为logistic regression. 📈

model <- aba_model() %>% 
  set_data(dat) %>% 
  set_groups(DX_bl %in% c('MCI','AD')) %>% 
  set_outcomes(ConvertedToAlzheimers, CSF_ABETA_STATUS_bl) %>% 
  set_predictors(
    PLASMA_PTAU181_bl,
    PLASMA_NFL_bl,
    c(PLASMA_PTAU181_bl, PLASMA_NFL_bl)
  ) %>% 
  set_covariates(AGE, GENDER, EDUCATION) %>% 
  set_stats(stat_glm(std.beta=T))

model

Note! 这里我们注意下如何进行Biomarker的联合应用, 可写为 c(PLASMA_PTAU181_bl, PLASMA_NFL_bl). 😘

5.3 拟合

model <- model %>% 
  fit()
model

5.4 模型数据

拟合完以后我们就获得了模型的结果, 大家可以大致看一下.🤓

model_summary <- model %>% 
  summary()

model_summary

6模型结果的可视化

这个包提供了很多可视化的参数, 可以直接将summary的结果传递给画图函数, 非常简单. 😘

6.1 coeffficients可视化

这里需要说明一下, 函数内自带的配色只有4种, 即jama, nature, lancet, none. 但是大家可以按照ggplot语法更改颜色.🤜

model_summary %>% 
  aba_plot_coef(coord_flip=T,
                palette = 'nature')

6.2 AUC可视化

看看各个模型的AUC吧.😗

model_summary %>% 
  aba_plot_metric(palette = 'nature')

6.3 ROC可视化

model_summary %>% 
  aba_plot_roc()

6.4 Risk density plot

批量出图, 大家想看哪个predictor的Risk density plot就提取哪个吧. 🤪

fig <- model %>%
  aba_plot_risk_density()
fig

随便提取一个看看吧~~~😉

fig$fig[1]

7补充一下

这个包还有很多其他强大的功能， 大家可以去进一步地探索一下。😘

欢迎大家留言说一下其他强大的函数哦~

这里附上所有函数官方解释：👇

• aba_adjust() Create an aba_adjust object.
• aba_control() Create an aba control object.
• aba_demographics() Create a demographics table from a fitted aba model.
• aba_diagnosticpower() Caclulate diagnostic power based on a fitted aba model
• aba_emmeans() Calculated estimated marginal means.
• aba_evaluate() Evaluate a fitted aba model on new data
• aba_fit() Fit an aba model.
• aba_longpower() Run power analysis on a longitudinal-based aba model.
• aba_model() Create an aba model.
• aba_plot() Plot an aba object
• aba_plot_coef() Plot coefficients of an aba model summary
• aba_plot_metric() Plot metrics of an aba model summary
• aba_plot_predictor_risk() Plot predictor values versus predicted risk from fitted aba model
• aba_plot_risk_density() Plot risk density split by binary outcome class
• aba_plot_roc() Plot ROC curves from an aba model
• aba_predict() Get individual predictions from a fitted aba model
• aba_read() Read an aba object from file
• aba_robust() Evaluate the robustness of an aba model to systematic and random error.
• aba_screen() Create an aba screen object.
• aba_selection() Run model selection on an aba model.
• aba_summary() Summarise a fitted aba model.
• aba_write() Write an aba object to file.
• adnimerge A sample of ADNI data in long format
• all_combos() Create all possible combinations of a set of variables
• all_levels() Create groups from all levels of one or more variables
• as_reactable() Convert an aba summary to a interactive react table
• as_reactable(*<abaSummary>*) Convert an aba summary to a interactive react table
• as_table() Convert an aba summary to a nicely formatted table
• as_table(*<abaSummary>*) Convert an aba summary to a nicely formatted table
• eval_boot() Create a bootstrap evaluator
• eval_cv() Create a cross validation evaluator
• eval_standard() Create a standard evaluator
• eval_traintest() Create a train-test evaluator
• everyone() Use all data rows as a group in an aba model.
• fit(*<abaModel>*) Fit an aba model.
• predict(*<abaModel>*) Get individual predictions from a fitted aba model
• set_covariates() Set the covariates of an aba model.
• set_data() Set the data of an aba model
• set_evals() Set the evals of an aba model
• set_groups() Set the groups of an aba model.
• set_outcomes() Set the outcomes of an aba model.
• set_predictors() Set the predictors of an aba model.
• set_stats() Set the stats of an aba model
• stat_ancova() Create an ancova stat object.
• stat_cox() Create a glm stat object.
• stat_glm() Create a glm stat object.
• stat_lm() Create an lm stat object.
• stat_lme() Create an lme stat object.
• stat_lmer() Create an lmer stat object.
• stat_mmrm() Create an mmrm stat object.
• stat_retest() Create a retest stat object.
• stat_roc() Create a roc stat object.
• theme_aba() Custom aba ggplot2 theme

最后祝大家早日不卷!~