Refactor analysis, Add Wilcoxon, Add ANOVA

README.md

@@ -60,7 +60,7 @@ docker build -t metabulo .
 and start the instance
 
 ```sh
-docker run -it -p 8004:8004 metabulo
+docker run -it --rm -p 8004:8004 metabulo
 ```
 
 You may find that changes to the position or size of the window in which the

devops/Dockerfile

@@ -1,6 +1,6 @@
 FROM metabulo/opencpu-base
 
 ADD metabulo /metabulo
-RUN R -e 'library(devtools); library(roxygen2); setwd("/metabulo"); document(); setwd("/"); install("metabulo")'
+RUN R -e 'library(devtools); library(roxygen2); setwd("/metabulo"); devtools::update_packages(devtools::dev_package_deps()$package, dependencies=NA, quiet=TRUE, upgrade="never"); document();  setwd("/");install("metabulo", dependencies=NA, quiet=TRUE, upgrade="never")'
 
 CMD service cron start && /usr/lib/rstudio-server/bin/rserver && apachectl -DFOREGROUND

devops/metabulo/DESCRIPTION

@@ -4,6 +4,7 @@ Version: 0.0.0.9000
 Authors@R: person("First", "Last", email = "first.last@example.com", role = c("aut", "cre"))
 Description: The package contains functions that are exposed via RPC for the Metabolomics web application.
 Depends: R (>= 3.5.2)
+Imports: car, emmeans
 License: Apache 2.0
 Encoding: UTF-8
 LazyData: true

devops/metabulo/R/analyses.R

@@ -0,0 +1,35 @@
+#' Basic PCA plot
+#'
+#' Uses MetaboAnalystR's PlotPCAPairSummary method
+
+anova_tukey_adjustment <- function(measurements, groups) {
+  Metab = read.csv(measurements, row.names=1)
+  groups = read.csv(groups, row.names=1)
+
+  # take the first column
+  Group = as.factor(groups[, 1])
+
+  library(car) #For Type III ANOVA
+  library(emmeans) #For pairwise comparisons https://cran.r-project.org/web/packages/emmeans/vignettes/comparisons.html
+
+  # based on the given R code
+  OUT <- NULL
+  for (n in colnames(Metab)){
+    mod <- lm(Metab[[n]] ~ Group)
+    a <- Anova(mod, type="III")
+    x1 <- cbind(n,t(a[,"Pr(>F)"] ))
+
+    emm.mod <- emmeans(mod, "Group")
+    b <- pairs(emm.mod)
+    x2 <- summary(b)$p.value
+    x2 <- t(x2)
+    colnames(x2) <- summary(b)$contrast
+
+    x <- cbind(x1,x2)
+    OUT <- rbind(OUT, x)
+    colnames(OUT) <- c("Metabolite", row.names(a), colnames(x2))
+    rm(mod,a,b,x1,x2,x)
+  }
+
+  OUT
+}

metabulo/analyses.py

@@ -0,0 +1,48 @@
+from itertools import combinations
+from typing import Any, Dict, List, Optional, Union
+
+import pandas as pd
+from scipy.stats import wilcoxon
+
+from .opencpu import opencpu_request
+
+
+def wilcoxon_test(measurements: pd.DataFrame, zero_method: Optional[str] = None,
+                  alternative: Optional[str] = None) -> Dict[str, Union[List[str], List[Dict]]]:
+
+    if zero_method is None:
+        zero_method = 'wilcox'
+    if alternative is None:
+        alternative = 'two-sided'
+
+    indices = list(measurements.index)
+    data = []
+
+    # test all possible combinations of rows
+    for (x, y) in combinations(measurements.iterrows(), 2):
+        x_index, x_data = x
+        y_index, y_data = y
+        (score, p) = wilcoxon(x_data, y_data, zero_method=zero_method, alternative=alternative)
+        data.append(dict(x=x_index, y=y_index, score=score, p=p))
+
+    return {
+        'indices': indices,
+        'data': data
+    }
+
+
+def anova_test(measurements: pd.DataFrame, groups: pd.Series) -> Dict[str, Any]:
+    files = {
+        'measurements': measurements.to_csv().encode(),
+        'groups': groups.to_csv(header=True).encode()
+    }
+
+    data = opencpu_request('anova_tukey_adjustment', files, {})
+
+    data = data.rename(columns={'(Intercept)': 'Intercept'})
+
+    return {
+        'groups': list(set(groups)),
+        'pairs': list(data)[4:],
+        'data': data.replace({pd.np.nan: None}).to_dict(orient='records')
+    }

metabulo/views.py

@@ -3,7 +3,7 @@
 import json
 import math
 from pathlib import PurePath
-from typing import cast, Dict
+from typing import cast, Dict, Optional
 from uuid import uuid4
 
 from flask import Blueprint, current_app, jsonify, request, Response, send_file
@@ -13,6 +13,7 @@
 from werkzeug import FileStorage
 
 from metabulo import opencpu
+from metabulo.analyses import anova_test, wilcoxon_test
 from metabulo.cache import csv_file_cache
 from metabulo.imputation import IMPUTE_MCAR_METHODS, IMPUTE_MNAR_METHODS
 from metabulo.models import AXIS_NAME_TYPES, CSVFile, CSVFileSchema, db, \
@@ -529,3 +530,39 @@ def get_pca_overview(validated_table):
     png_content = opencpu.generate_image('/metabulo/R/pca_overview_plot',
                                          validated_table.measurements)
     return Response(png_content, mimetype='image/png')
+
+
+@csv_bp.route('/csv/<uuid:csv_id>/analyses/wilcoxon', methods=['GET'])
+@use_kwargs({
+    'zero_method': fields.Str(missing='wilcox',
+                              validate=validate.OneOf(['wilcox', 'pratt', 'zsplit'])),
+    'alternative': fields.Str(missing='two-sided',
+                              validate=validate.OneOf(['two-sided', 'greater', 'less']))
+})
+@load_validated_csv_file
+def get_wilcoxon_test(validated_table: ValidatedMetaboliteTable,
+                      zero_method: str, alternative: str):
+    table = validated_table.measurements
+
+    data = wilcoxon_test(table, zero_method, alternative)
+
+    return jsonify(data), 200
+
+
+@csv_bp.route('/csv/<uuid:csv_id>/analyses/anova', methods=['GET'])
+@use_kwargs({
+    'group_column': fields.Str()
+})
+@load_validated_csv_file
+def get_anova_test(validated_table: ValidatedMetaboliteTable, group_column: Optional[str] = None):
+    measurements = validated_table.measurements
+    groups = validated_table.groups
+
+    group = groups.iloc[:, 0] if group_column is None else groups.loc[:, group_column]
+    if group is None:
+        raise ValidationError(
+            'invalid group column', field_name='group_column', data=group_column)
+
+    data = anova_test(measurements, group)
+
+    return jsonify(data), 200

web/src/common/api.service.js

@@ -51,6 +51,10 @@ export const CSVService = {
     return ApiService.get('csv', `${slug}/plot/${type}`, params);
   },
 
+  getAnalysis(slug, key, params = {}) {
+    return ApiService.get('csv', `${slug}/analyses/${key}`, params);
+  },
+
   post(data) {
     return ApiService.post('csv', data);
   },

web/src/components/AnalyzeData.vue

@@ -0,0 +1,35 @@
+<script>
+import Vue from 'vue';
+import { analyses } from './vis';
+
+export default Vue.extend({
+  props: {
+    id: {
+      type: String,
+      required: true,
+    },
+  },
+  data() {
+    return {
+      analyses,
+    };
+  },
+});
+</script>
+
+<template lang="pug">
+v-layout.analyze-component(row, fill-height)
+  v-container.grow-overflow.ma-0(grid-list-lg, fluid)
+    v-container.pa-2(fluid)
+      v-layout(row)
+        v-flex(v-for="card in analyses", :key="card.path", md3)
+          v-card
+            v-card-title(primary-title)
+              div
+                .headline(v-text="card.name")
+                | {{card.description}}
+            v-card-actions
+              v-btn(text,
+                  @click="$router.push({ path: `/pretreatment/${id}/analyze/${card.path}` })")
+                | Analyze
+</template>

web/src/components/Pretreatment.vue

@@ -1,6 +1,7 @@
 <script>
 import { SET_SELECTION } from '@/store/mutations.type';
 import { loadDataset } from '@/utils/mixins';
+import { analyses } from './vis';
 
 export default {
   mixins: [loadDataset],
@@ -13,6 +14,7 @@ export default {
   data() {
     return {
       datasets: this.$store.state.datasets,
+      analyses,
     };
   },
   methods: {
@@ -42,6 +44,9 @@ export default {
         });
       }
     },
+    isSubRoute(dataset, start) {
+      return this.$router.currentRoute.path.startsWith(`/pretreatment/${dataset.id}/${start}`);
+    },
   },
 };
 </script>
@@ -95,6 +100,26 @@ v-layout.pretreatment-component(row, fill-height)
               v-list-tile-title.pl-2
                 v-icon.pr-1.middle {{ $vuetify.icons.bubbles }}
                 | Transform Table
+
+            v-list-group.ml-2.link-group(
+                :append-icon="null",
+                :value="isSubRoute(dataset, `analyze/`)",
+                :disabled="!valid(dataset)",
+                @click="$router.push({ path: `/pretreatment/${dataset.id}/analyze` })")
+              template(v-slot:activator)
+                v-list-tile(
+                   :class="{ active: $router.currentRoute.name === 'Analyze Data' }")
+                  v-list-tile-title.pl-2
+                    v-icon.pr-1.middle {{ $vuetify.icons.cogs }}
+                    | Analyze Table
+              v-list-tile.ml-2.small-tile(
+                  v-for="a in analyses", :key="a.path",
+                  :class="{ active: $router.currentRoute.name === a.shortName }",
+                  @click="$router.push({ path: `/pretreatment/${dataset.id}/analyze/${a.path}` })")
+                v-list-tile-title.pl-2
+                  v-icon.pr-1.middle {{ $vuetify.icons.compare }}
+                  | {{a.shortName}}
+
   keep-alive
     router-view
 </template>
@@ -125,7 +150,8 @@ v-layout.pretreatment-component(row, fill-height)
     }
   }
 
-  .view-list .v-list__tile--link {
+  .view-list .v-list__tile--link,
+  .view-list .link-group > .v-list__group__header {
     transition: none;
 
     .v-list__tile__title {

web/src/components/ToolbarOption.vue

@@ -0,0 +1,31 @@
+<script>
+export default {
+  props: {
+    title: {
+      type: String,
+      required: true,
+    },
+    value: {
+      type: String,
+      required: true,
+    },
+    options: {
+      type: Array,
+      required: true,
+    },
+  },
+};
+</script>
+
+<template lang="pug">
+div
+  v-toolbar.darken-3(color="primary", dark, flat, dense, :card="false")
+    v-toolbar-title {{title}}
+  v-card.mx-3(flat)
+    v-card-actions
+      v-radio-group.my-0(:value="value",
+          hide-details,
+          @change="$emit('change', $event)")
+        v-radio(v-for="m in options", :label="m.label",
+            :value="m.value", :key="m.value")
+</template>

web/src/components/Transform.vue

@@ -1,6 +1,6 @@
 <script>
 import { mapState } from 'vuex';
-import { MUTEX_TRANSFORM_TABLE, LOAD_PLOT } from '@/store/actions.type';
+import { MUTEX_TRANSFORM_TABLE } from '@/store/actions.type';
 import {
   normalize_methods,
   scaling_methods,
@@ -39,46 +39,8 @@ export default {
     norm_arg() { return this.$store.getters.txType(this.id, 'normalization_argument'); },
     trans() { return this.$store.getters.txType(this.id, 'transformation'); },
     scaling() { return this.$store.getters.txType(this.id, 'scaling'); },
-    pcaData() { return this.$store.getters.plotData(this.id, 'pca'); },
-    pcaValid() { return this.$store.getters.plotValid(this.id, 'pca'); },
-    loadingsData() { return this.$store.getters.plotData(this.id, 'loadings'); },
-    loadingsValid() { return this.$store.getters.plotValid(this.id, 'loadings'); },
-  },
-  watch: {
-    pcaValid: {
-      immediate: true,
-      handler(valid) { this.pcaLoader(valid); },
-    },
-    loadingsValid: {
-      immediate: true,
-      handler(valid) { this.loadingsLoader(valid); },
-    },
-    id() {
-      this.pcaLoader(this.pcaValid);
-      this.loadingsLoader(this.loadingsValid);
-    },
-    loading() {
-      this.pcaLoader(this.pcaValid);
-      this.loadingsLoader(this.loadingsValid);
-    },
   },
   methods: {
-    pcaLoader(valid) {
-      if (valid === false) {
-        this.$store.dispatch(LOAD_PLOT, {
-          dataset_id: this.id,
-          name: 'pca',
-        });
-      }
-    },
-    loadingsLoader(valid) {
-      if (valid === false) {
-        this.$store.dispatch(LOAD_PLOT, {
-          dataset_id: this.id,
-          name: 'loadings',
-        });
-      }
-    },
     async transformTable(value, category, argument, methods) {
       /* If there's no argument and there should be, pick the firt from the list */
       const method = methods.find(v => v.value === value);
@@ -157,16 +119,15 @@ v-layout.transform-component(row, fill-height)
         score-plot-tile(
             :width="600",
             :height="600",
-            :raw-points="pcaData",
-            :dataset="dataset")
+            :id="id")
         loadings-plot-tile(
             :width="600",
             :height="600",
-            :points="loadingsData")
+            :id="id")
         scree-plot-tile(
             :width="600",
             :height="600",
-            :eigenvalues="pcaData.sdev")
+            :id="id")
   v-container(v-else-if="ready", fill-height)
     v-layout(column)
       .display-2 Error: Cannot show transform table

web/src/components/stepper/Stepper.vue

@@ -18,7 +18,7 @@ const step3 = {
 };
 const step4 = {
   name: 'Analysis',
-  description: '',
+  description: 'Analyze your data using statistical tests and visualizations',
   icon: 'checkedFlag',
 };
 const steps = [step1, step2, step3, step4];