add missing test file

Author: cokelaer

test/regression/test_regression_report.py

@@ -0,0 +1,95 @@
+from gdsctools import *
+import pandas as pd
+import pylab
+import json
+
+def test_regression_report():
+    IC = gdsctools_data("IC50_v5.csv.gz")
+    GF = gdsctools_data("genomic_features_v5.csv.gz")
+
+    PREFIX = "gdsctools_regression_"
+    IMAGE_DIR = "images"
+    DATA_PREFIX = "data/" + PREFIX
+    IMAGE_PREFIX = "images/" + PREFIX
+
+    gd = regression.GDSCLasso(IC, GF)
+    DRUGIDS = gd.drugIds[0:4]
+
+    config = {"boxplot_n":5, "randomness":5}
+
+    # Get best model
+    inputs = []
+    for drugid in DRUGIDS:
+        res = gd.runCV(drugid, verbose=False, kfolds=10)
+        bestmodel = gd.get_model(alpha=res.alpha)
+
+        def _pngname(tag):
+            return IMAGE_PREFIX + "%s_%s.png" % (tag, drugid)
+
+        # Plot weights
+        weights = gd.plot_weight(drugid, bestmodel)
+        if len(weights):
+            pylab.savefig(_pngname("weights"))
+            pylab.close()
+
+        weights = pd.DataFrame({
+                "weigths": res.coefficients,
+                "features":gd.feature_names})
+        output = DATA_PREFIX + "weights_{}.csv".format(drugid)
+        weights.to_csv(output, index=False)
+
+        # Plot importance
+        weights = gd.plot_importance(drugid, bestmodel)
+        if len(weights):
+            pylab.savefig(_pngname("importance"))
+            pylab.close()
+
+        # Boxplots
+        boxres = gd.boxplot(drugid, model=bestmodel, n=5,
+            bx_vert=False)
+        if len(boxres['data']):
+            pylab.savefig(_pngname("boxplot"))
+            pylab.close()
+
+        # Bayes factor
+        ran = gd.check_randomness(drugid, 10, 10)
+        pylab.savefig(_pngname("randomness"))
+        pylab.close()
+        results = {"drugid": int(drugid),
+                "Rp":res.Rp,
+                "alpha": res.alpha,
+                "ln_alpha": res.ln_alpha,
+                "ttest": ran['ttest_pval'],
+                "bayes":ran['bayes_factor']}
+
+        output = DATA_PREFIX + "results_{}.json".format(drugid)
+        fh = open(output, "w")
+        json.dump(results, fh)
+        fh.close()
+        inputs.append(output)
+
+    # gather all results:
+    data = []
+    for this in inputs:
+        with open(this, "r") as fh:
+            data.append(json.loads(fh.read()))
+    df = pd.DataFrame(data)
+    df.set_index("drugid", inplace=True)
+    df.to_csv(DATA_PREFIX + "results.csv")
+
+
+    inputs = [DATA_PREFIX + "weights_{}.csv".format(drugid) for drugid in DRUGIDS]
+    df = pd.concat(
+       [pd.read_csv(this).set_index("features") for this in inputs],
+            axis=1)
+    df.columns = DRUGIDS
+    df.to_csv(DATA_PREFIX + "weights.csv")
+
+
+    from gdsctools import regression_report
+    report = regression_report.RegressionReport("lasso", image_dir=IMAGE_DIR, config=config)
+    report.create_html_main()
+    report.create_html_drug()
+
+
+