Merge pull request #565 from VEuPathDB/fix-528-volcano-point-floors

Incorporate p value floor from backend
VEuPathDB · Oct 30, 2023 · a8a7b4f · a8a7b4f
2 parents 38c98c7 + a22951a
commit a8a7b4f
Show file tree

Hide file tree

Showing 5 changed files with 128 additions and 37 deletions.
diff --git a/packages/libs/components/src/plots/VolcanoPlot.tsx b/packages/libs/components/src/plots/VolcanoPlot.tsx
@@ -46,6 +46,17 @@ export interface RawDataMinMaxValues {
   y: NumberRange;
 }
 
+export interface StatisticsFloors {
+  /** The minimum allowed p value. Useful for protecting the plot against taking the log of pvalue=0. Points with true pvalue <= pValueFloor will get plotted at -log10(pValueFloor).
+   * Any points with pvalue <= the pValueFloor will show "P Value <= {pValueFloor}" in the tooltip.
+   */
+  pValueFloor: number;
+  /** The minimum allowed adjusted p value. Ideally should be calculated in conjunction with the pValueFloor. Currently used
+   * only to update the tooltip with this information, but later will be used to control the y axis location, similar to pValueFloor.
+   */
+  adjustedPValueFloor?: number;
+}
+
 export interface VolcanoPlotProps {
   /** Data for the plot. An effectSizeLabel and an array of VolcanoPlotDataPoints */
   data: VolcanoPlotData | undefined;
@@ -80,8 +91,10 @@ export interface VolcanoPlotProps {
   showSpinner?: boolean;
   /** used to determine truncation logic */
   rawDataMinMaxValues: RawDataMinMaxValues;
-  /** The maximum possible y axis value. Points with pValue=0 will get plotted at -log10(minPValueCap). */
-  minPValueCap?: number;
+  /** Minimum (floor) values for p values and adjusted p values. Will set a cap on the maximum y axis values
+   * at which points can be plotted. This information will also be shown in tooltips for floored points.
+   */
+  statisticsFloors?: StatisticsFloors;
 }
 
 const EmptyVolcanoPlotStats: VolcanoPlotStats = [
@@ -93,6 +106,10 @@ const EmptyVolcanoPlotData: VolcanoPlotData = {
   statistics: EmptyVolcanoPlotStats,
 };
 
+export const DefaultStatisticsFloors: StatisticsFloors = {
+  pValueFloor: 0, // Do not floor by default
+};
+
 const MARGIN_DEFAULT = 50;
 
 interface TruncationRectangleProps {
@@ -144,7 +161,7 @@ function VolcanoPlot(props: VolcanoPlotProps, ref: Ref<HTMLDivElement>) {
     truncationBarFill,
     showSpinner = false,
     rawDataMinMaxValues,
-    minPValueCap = 2e-300,
+    statisticsFloors = DefaultStatisticsFloors,
   } = props;
 
   // Use ref forwarding to enable screenshotting of the plot for thumbnail versions.
@@ -167,34 +184,44 @@ function VolcanoPlot(props: VolcanoPlotProps, ref: Ref<HTMLDivElement>) {
   const { min: dataXMin, max: dataXMax } = rawDataMinMaxValues.x;
   const { min: dataYMin, max: dataYMax } = rawDataMinMaxValues.y;
 
+  // When dataYMin = 0, there must be a point with pvalue = 0, which means the plot will try in vain to draw a point at -log10(0) = Inf.
+  // When this issue arises, one should set a pValueFloor >= 0 so that the point with pValue = 0
+  // will be able to be plotted sensibly.
+  if (dataYMin === 0 && statisticsFloors.pValueFloor <= 0) {
+    throw new Error(
+      'Found data point with pValue = 0. Cannot create a volcano plot with a point at -log10(0) = Inf. Please use the statisticsFloors prop to set a pValueFloor >= 0.'
+    );
+  }
+
   // Set mins, maxes of axes in the plot using axis range props
-  // The y axis max should not be allowed to exceed -log10(minPValueCap)
+  // The y axis max should not be allowed to exceed -log10(pValueFloor)
   const xAxisMin = independentAxisRange?.min ?? 0;
   const xAxisMax = independentAxisRange?.max ?? 0;
   const yAxisMin = dependentAxisRange?.min ?? 0;
   const yAxisMax = dependentAxisRange?.max
-    ? dependentAxisRange.max > -Math.log10(minPValueCap)
-      ? -Math.log10(minPValueCap)
+    ? dependentAxisRange.max > -Math.log10(statisticsFloors.pValueFloor)
+      ? -Math.log10(statisticsFloors.pValueFloor)
       : dependentAxisRange.max
     : 0;
 
   // Do we need to show the special annotation for the case when the y axis is maxxed out?
-  const showCappedDataAnnotation = yAxisMax === -Math.log10(minPValueCap);
+  const showFlooredDataAnnotation =
+    yAxisMax === -Math.log10(statisticsFloors.pValueFloor);
 
   // Truncation indicators
   // If we have truncation indicators, we'll need to expand the plot range just a tad to
   // ensure the truncation bars appear. The folowing showTruncationBar variables will
   // be either 0 (do not show bar) or 1 (show bar).
-  // The y axis has special logic because it gets capped at -log10(minPValueCap)
+  // The y axis has special logic because it gets capped at -log10(pValueFloor) and we dont want to
+  // show the truncation bar if the annotation will be shown!
   const showXMinTruncationBar = Number(dataXMin < xAxisMin);
   const showXMaxTruncationBar = Number(dataXMax > xAxisMax);
   const xTruncationBarWidth = 0.02 * (xAxisMax - xAxisMin);
 
   const showYMinTruncationBar = Number(-Math.log10(dataYMax) < yAxisMin);
-  const showYMaxTruncationBar =
-    dataYMin === 0
-      ? Number(-Math.log10(minPValueCap) > yAxisMax)
-      : Number(-Math.log10(dataYMin) > yAxisMax);
+  const showYMaxTruncationBar = Number(
+    -Math.log10(dataYMin) > yAxisMax && !showFlooredDataAnnotation
+  );
   const yTruncationBarHeight = 0.02 * (yAxisMax - yAxisMin);
 
   /**
@@ -211,12 +238,12 @@ function VolcanoPlot(props: VolcanoPlotProps, ref: Ref<HTMLDivElement>) {
    * Accessors - tell visx which value of the data point we should use and where.
    */
 
-  // For the actual volcano plot data. Y axis points are capped at -Math.log10(minPValueCap)
+  // For the actual volcano plot data. Y axis points are capped at -Math.log10(pValueFloor)
   const dataAccessors = {
     xAccessor: (d: VolcanoPlotDataPoint) => Number(d?.effectSize),
     yAccessor: (d: VolcanoPlotDataPoint) =>
-      d.pValue === '0'
-        ? -Math.log10(minPValueCap)
+      Number(d.pValue) <= statisticsFloors.pValueFloor
+        ? -Math.log10(statisticsFloors.pValueFloor)
         : -Math.log10(Number(d?.pValue)),
   };
 
@@ -266,7 +293,7 @@ function VolcanoPlot(props: VolcanoPlotProps, ref: Ref<HTMLDivElement>) {
           findNearestDatumOverride={findNearestDatumXY}
           margin={{
             top: MARGIN_DEFAULT,
-            right: showCappedDataAnnotation ? 150 : MARGIN_DEFAULT,
+            right: showFlooredDataAnnotation ? 150 : MARGIN_DEFAULT,
             left: MARGIN_DEFAULT,
             bottom: MARGIN_DEFAULT,
           }}
@@ -351,7 +378,7 @@ function VolcanoPlot(props: VolcanoPlotProps, ref: Ref<HTMLDivElement>) {
           )}
 
           {/* infinity y data annotation line */}
-          {showCappedDataAnnotation && (
+          {showFlooredDataAnnotation && (
             <Annotation
               datum={{
                 x: xAxisMax,
@@ -441,11 +468,23 @@ function VolcanoPlot(props: VolcanoPlotProps, ref: Ref<HTMLDivElement>) {
                       <span>{effectSizeLabel}:</span> {data?.effectSize}
                     </li>
                     <li>
-                      <span>P Value:</span> {data?.pValue}
+                      <span>P Value:</span>{' '}
+                      {data?.pValue
+                        ? Number(data.pValue) <= statisticsFloors.pValueFloor
+                          ? '<= ' + statisticsFloors.pValueFloor
+                          : data?.pValue
+                        : 'n/a'}
                     </li>
                     <li>
                       <span>Adjusted P Value:</span>{' '}
-                      {data?.adjustedPValue ?? 'n/a'}
+                      {data?.adjustedPValue
+                        ? statisticsFloors.adjustedPValueFloor &&
+                          Number(data.adjustedPValue) <=
+                            statisticsFloors.adjustedPValueFloor &&
+                          Number(data.pValue) <= statisticsFloors.pValueFloor
+                          ? '<= ' + statisticsFloors.adjustedPValueFloor
+                          : data?.adjustedPValue
+                        : 'n/a'}
                     </li>
                   </ul>
                 </div>

diff --git a/packages/libs/components/src/stories/plots/VolcanoPlot.stories.tsx b/packages/libs/components/src/stories/plots/VolcanoPlot.stories.tsx
@@ -1,4 +1,7 @@
-import VolcanoPlot, { VolcanoPlotProps } from '../../plots/VolcanoPlot';
+import VolcanoPlot, {
+  StatisticsFloors,
+  VolcanoPlotProps,
+} from '../../plots/VolcanoPlot';
 import { Story, Meta } from '@storybook/react/types-6-0';
 import { range } from 'lodash';
 import { getNormallyDistributedRandomNumber } from './ScatterPlot.storyData';
@@ -70,8 +73,8 @@ const dataSetVolcano: VEuPathDBVolcanoPlotData = {
         '0.001',
         '0.0001',
         '0.002',
-        '0',
-        '0',
+        '1e-90',
+        '0.00000002',
       ],
       adjustedPValue: ['0.01', '0.001', '0.01', '0.001', '0.02', '0', '0'],
       pointID: [
@@ -123,6 +126,7 @@ interface TemplateProps {
   comparisonLabels?: string[];
   truncationBarFill?: string;
   showSpinner?: boolean;
+  statisticsFloors?: StatisticsFloors;
 }
 
 const Template: Story<TemplateProps> = (args) => {
@@ -131,22 +135,27 @@ const Template: Story<TemplateProps> = (args) => {
   const volcanoDataPoints: VolcanoPlotData | undefined = {
     effectSizeLabel: args.data?.volcanoplot.effectSizeLabel ?? '',
     statistics:
-      args.data?.volcanoplot.statistics.effectSize.map((effectSize, index) => {
-        return {
-          effectSize: effectSize,
-          pValue: args.data?.volcanoplot.statistics.pValue[index],
-          adjustedPValue:
-            args.data?.volcanoplot.statistics.adjustedPValue[index],
-          pointID: args.data?.volcanoplot.statistics.pointID[index],
+      args.data?.volcanoplot.statistics.effectSize
+        .map((effectSize, index) => {
+          return {
+            effectSize: effectSize,
+            pValue: args.data?.volcanoplot.statistics.pValue[index],
+            adjustedPValue:
+              args.data?.volcanoplot.statistics.adjustedPValue[index],
+            pointID: args.data?.volcanoplot.statistics.pointID[index],
+          };
+        })
+        .map((d) => ({
+          ...d,
+          pointIDs: d.pointID ? [d.pointID] : undefined,
           significanceColor: assignSignificanceColor(
-            Number(effectSize),
-            Number(args.data?.volcanoplot.statistics.pValue[index]),
+            Number(d.effectSize),
+            Number(d.pValue),
             args.significanceThreshold,
             args.effectSizeThreshold,
             significanceColors
           ),
-        };
-      }) ?? [],
+        })) ?? [],
   };
 
   const rawDataMinMaxValues = {
@@ -191,6 +200,7 @@ const Template: Story<TemplateProps> = (args) => {
     truncationBarFill: args.truncationBarFill,
     showSpinner: args.showSpinner,
     rawDataMinMaxValues,
+    statisticsFloors: args.statisticsFloors,
   };
 
   return (
@@ -268,3 +278,21 @@ Empty.args = {
   independentAxisRange: { min: -9, max: 9 },
   dependentAxisRange: { min: -1, max: 9 },
 };
+
+// With a pvalue floor
+const testStatisticsFloors: StatisticsFloors = {
+  pValueFloor: 0.006,
+  adjustedPValueFloor: 0.01,
+};
+export const FlooredPValues = Template.bind({});
+FlooredPValues.args = {
+  data: dataSetVolcano,
+  markerBodyOpacity: 0.8,
+  effectSizeThreshold: 1,
+  significanceThreshold: 0.01,
+  comparisonLabels: ['up in group a', 'up in group b'],
+  independentAxisRange: { min: -9, max: 9 },
+  dependentAxisRange: { min: -1, max: 9 },
+  showSpinner: false,
+  statisticsFloors: testStatisticsFloors,
+};
diff --git a/packages/libs/eda/src/lib/core/api/DataClient/types.ts b/packages/libs/eda/src/lib/core/api/DataClient/types.ts
@@ -366,10 +366,16 @@ export const VolcanoPlotStatistics = array(
   })
 );
 
-export const VolcanoPlotResponse = type({
-  effectSizeLabel: string,
-  statistics: VolcanoPlotStatistics,
-});
+export const VolcanoPlotResponse = intersection([
+  type({
+    effectSizeLabel: string,
+    statistics: VolcanoPlotStatistics,
+  }),
+  partial({
+    pValueFloor: string,
+    adjustedPValueFloor: string,
+  }),
+]);
 
 export interface VolcanoPlotRequestParams {
   studyId: string;

diff --git a/packages/libs/eda/src/lib/core/components/computations/plugins/differentialabundance.tsx b/packages/libs/eda/src/lib/core/components/computations/plugins/differentialabundance.tsx
@@ -73,6 +73,7 @@ export const DifferentialAbundanceConfig = t.type({
   collectionVariable: VariableCollectionDescriptor,
   comparator: Comparator,
   differentialAbundanceMethod: t.string,
+  pValueFloor: t.string,
 });
 
 // Check to ensure the entirety of the configuration is filled out before enabling the
@@ -188,6 +189,11 @@ export function DifferentialAbundanceConfiguration(
     configuration.differentialAbundanceMethod =
       DIFFERENTIAL_ABUNDANCE_METHODS[0];
 
+  // Set the pValueFloor here. May change for other apps.
+  // Note this is intentionally different than the default pValueFloor used in the Volcano component. By default
+  // that component does not floor the data, but we know we want the diff abund computation to use a floor.
+  if (configuration) configuration.pValueFloor = '1e-200';
+
   // Include known collection variables in this array.
   const collections = useCollectionVariables(studyMetadata.rootEntity);
   if (collections.length === 0)

diff --git a/...s/eda/src/lib/core/components/visualizations/implementations/VolcanoPlotVisualization.tsx b/...s/eda/src/lib/core/components/visualizations/implementations/VolcanoPlotVisualization.tsx
@@ -3,6 +3,8 @@ import VolcanoPlot, {
   VolcanoPlotProps,
   assignSignificanceColor,
   RawDataMinMaxValues,
+  StatisticsFloors,
+  DefaultStatisticsFloors,
 } from '@veupathdb/components/lib/plots/VolcanoPlot';
 
 import * as t from 'io-ts';
@@ -415,6 +417,15 @@ function VolcanoPlotViz(props: VisualizationProps<Options>) {
         ]
       : [];
 
+  // Record any floors for the p value and adjusted p value sent to us from the backend.
+  const statisticsFloors: StatisticsFloors =
+    data.value && data.value.pValueFloor
+      ? {
+          pValueFloor: Number(data.value.pValueFloor),
+          adjustedPValueFloor: Number(data.value.adjustedPValueFloor),
+        }
+      : DefaultStatisticsFloors;
+
   const volcanoPlotProps: VolcanoPlotProps = {
     /**
      * VolcanoPlot defines an EmptyVolcanoPlotData variable that will be assigned when data is undefined.
@@ -437,6 +448,7 @@ function VolcanoPlotViz(props: VisualizationProps<Options>) {
      * confusing behavior where selecting group values displays on the empty viz placeholder.
      */
     comparisonLabels: data.value ? comparisonLabels : [],
+    statisticsFloors,
     showSpinner: data.pending,
     truncationBarFill: yellow[300],
     independentAxisRange,