@datagrok/curves NPM

Curves

Curves provides support for fitted curves (such as dose-response curves), including in-grid rendering, storing charts in cells, interactivity, and automatic fitting.

Fitting: computing parameters of the specified function to best fit the data
- Uses the BFGS optimization algorithm (multi-threading for performance). For dose-response curves, we are typically fitting the sigmoid function
- Ability to dynamically register custom fitting functions
  - Automatic fit function determination
  - Caching of custom fitting functions
- Ability to get fitting performance characteristics (r-squared, classification, etc.)
Deep integration with the Datagrok grid
- Either fitting on the fly or using the supplied function + parameters
- Multiple series in one cell
- Candlesticks, confidence intervals, standard deviation and droplines drawing
- Ability to define chart, marker, or fitting options (such as fit function or marker color) on the column level, with the ability to override it on a grid cell or point level
- Clicking a point in a chart within a grid makes it an outlier -> curve is re-fitted on the fly
- Ability to switch curves parameters from the property panel on different levels (dataframe, column, cell)
- Ability to specify a chart as a "reference" so that it is shown on every other chart for comparison
Ability to overlay curves from multiple grid cells (special viewer)
Work with series stored in multiple formats (binary for performance, JSON for flexibility, etc.)

Rendering fitted curves using JSON format

To render a fitted curve based on series points, you need to write it in the following JSON format:

{
  "series": [
    {
      "name": "Test series",
      "pointColor": "#2ca02c",
      "fitLineColor": "#2ca02c",
      "confidenceIntervalColor": "#fbec5d",
      "outlierColor": "#d62728",
      "markerType": "circle",
      "lineStyle": "dashed",
      "errorModel": "proportional",
      "connectDots": false,
      "showFitLine": true,
      "showCurveConfidenceInterval": true,
      "fitFunction": "sigmoid",
      "parameters": [1.7391934768969721, -0.9451759934029763, 4.846020678949615, 0.15841886339211816],
      "parameterBounds": [
        {"min": 1.739193476892, "max": 1.739193476898},
        {"max": -0.9451759934029},
        {},
        {"min": 0.158418863392114}
      ],
      "showPoints": "points",
      "clickToToggle": true,
      "droplines": ["IC50"],
      "points": [
        { "x": 0.10000000149011612, "y": 0.04152340441942215, "stdev": 0.031523404876, "marker": "square" },
        { "x": 0.6000000238418579, "y": 0.11901605129241943, "outlier": true, "outlierColor": "#ff7f0e" },
        { "x": 1.100000023841858, "y": 0.11143334954977036, "outlier": false, "color": "#2ca02c", "size": 5 },
        // ...
      ]
    }
  ],
  "chartOptions": {
    "showStatistics": ["auc", "rSquared"],
    "minX": 0.10000000149011612,
    "minY": 0.04152340441942215,
    "maxX": 7.099999904632568,
    "maxY": 1.7591952085494995,
    "title": "Dose-Response curves",
    "xAxisName": "Concentration",
    "yAxisName": "Activity",
    "logX": true,
    "logY": false,
    "allowXZeroes": true,
    "mergeSeries": false,
    "showColumnLabel": true
  }
}

Each series has its own parameters, such as:

name - controls the series name
pointColor - overrides the standardized series point color
fitLineColor - overrides the standardized series fit line color
confidenceIntervalColor - overrides the standardized series confidence interval color
outlierColor - overrides the standardized series outlier color
markerType - defines the series marker type, which could be circle, asterisk, square, etc.
lineStyle - defines the series line style, which could be solid, dotted, dashed or dashdotted
errorModel - defines the series error model, which could be either constant or proportional
connectDots - defines whether to connect the points with lines or not
showFitLine - defines whether to show the fit line or not
showCurveConfidenceInterval - defines whether to show the confidence intervals or not
fitFunction - controls the series fit function, which could be either a sigmoid function or a custom-defined function.
parameters - controls the series parameters, if set explicitly - the fitting process won't be executed. The parameter order of the sigmoid function is: max, tan, IC50, min.
parameterBounds - defines the acceptable range of each parameter, which is taken into account during the fitting. See also parameters
showPoints - defines the data display mode, which could be either points, candlesticks, both, or none
clickToToggle - defines whether clicking on the point toggles its outlier status and causes curve refitting or not
droplines - defines the droplines that would be shown on the plot (for instance, IC50)
points - an array of objects with each object containing x and y coordinates and its own parameters:
- outlier - if true, renders as 'x' and gets ignored for curve fitting
- color - overrides the marker color defined in series pointColor
- marker - overrides the marker type defined in series markerType
- size - overrides the default marker size
- stdev - when defined, renders an error bar candlestick

Each chart has its own parameters as well, such as:

minX, minY, maxX, maxY - controls the minimum x and y values of the plot
title - defines the plot title. If the plot size is enough, will render it
xAxisName, yAxisName - defines the x and y axis names. If the plot size is enough, will render it
logX, logY - defines whether the x and y data should be logarithmic or not
allowXZeroes - defines whether x zeroes allowed for logarithmic data or not. If the flag is true, it will calculate the approximate log(0) that will fit the chart
mergeSeries - defines whether to merge series or not
showColumnLabel - defines whether to show the column label in the legend or not
showStatistics - defines the statistics that would be shown on the plot (such as the area under the curve (auc) or the coefficient of determination (rSquared))

curves

Creating a custom fit function

To render a custom fit function, you need to write in the following JSON format:

"fitFunction": {
  "name": "Polynomial",
  "function": "([p1, p2, p3, p4], x) => p1 * x * x * x + p2 * x * x + p3 * x + p4",
  "getInitialParameters": "(xs, ys) => [0.1, -1, 4, 4]",
  "parameterNames": ["Slope", "Intercept", "Parameter 3", "Parameter 4"]
}

Each fitting function has its own name, which is used to cache the created custom function, enabling efficient retrieval and reuse, and parameterNames, which are stored as an array of strings.

Also, there are two functions: getInitialParameters, which takes arrays of x and y and returns determined initial parameter values, and function, which takes the array of parameters and the given x coordinate and returns the result of the fit function. These functions are written as JavaScript arrow function expressions.

custom-fit-function

Multi Curve Viewer

You can overlay curves from multiple cells on one chart with the Multi Curve Viewer:

npm.io