comparison lifelines_tool/lifelineskmcph.xml @ 2:dd5e65893cb8 draft default tip

add survival and collapsed life table outputs suggested by Wolfgang

1:232b874046a7

<tool name="lifelineskmcph" id="lifelineskmcph" version="0.01">
  <!--Source in git at: https://github.com/fubar2/galaxy_tf_overlay-->
  <!--Created by [e-mail] at 10/08/2023 15:48:43 using the Galaxy Tool Factory.-->
  <description>Lifelines KM and optional Cox PH models</description>
  <requirements>
    <requirement version="1.5.3" type="package">pandas</requirement>
    <requirement version="3.7.2" type="package">matplotlib</requirement>
    <requirement version="0.27.7" type="package">lifelines</requirement>

    if colsok:
        df.columns = testcols # use actual header
    else:
        colsok = (args.time in defaultcols) and (args.status in defaultcols)
        if colsok:
            sys.stderr.write('replacing first row of data derived header %s with %s' % (testcols, defaultcols))
            df.columns = defaultcols
        else:
            sys.stderr.write('## CRITICAL USAGE ERROR (not a bug!): time %s and status %s do not match anything in the file header, supplied header or automatic default column names %s' % (args.time, args.status, defaultcols))
print('## Lifelines tool starting.\nUsing data header =', df.columns, 'time column =', args.time, 'status column =', args.status)
os.makedirs(args.image_dir, exist_ok=True)
fig, ax = plt.subplots()
if args.group > '':
    names = []
    times = []

    kmf.fit(df[args.time], df[args.status])
    kmf.plot_survival_function(ax=ax)
    ax.set_title(args.title)
    fig.savefig(os.path.join(args.image_dir,'KM_%s.png' % args.title))
    print('#### No grouping variable, so no log rank or other Kaplan-Meier statistical output is available')
if len(args.cphcols) > 0:
    fig, ax = plt.subplots()
    ax.set_title('Cox-PH model: %s' % args.title)
    cphcols = args.cphcols.strip().split(',')
    cphcols = [x.strip() for x in cphcols]

    cphdf = df[cutcphcols]
    ucolcounts = colsdf.nunique(axis=0)
    cph.fit(cphdf, duration_col=args.time, event_col=args.status)
    cph.print_summary()
    for i, cov in enumerate(colsdf.columns):
         if ucolcounts[i] > 10:
             v = pd.Series.tolist(cphdf[cov].quantile(QVALS))
             vdt = df.dtypes[cov]
             if vdt == 'int64':
                 v = trimlegend(v)
             axp = cph.plot_partial_effects_on_outcome(cov, cmap='coolwarm', values=v)

2:dd5e65893cb8

<tool name="lifelineskmcph" id="lifelineskmcph" version="0.01">
  <!--Source in git at: https://github.com/fubar2/galaxy_tf_overlay-->
  <!--Created by [e-mail] at 10/08/2023 21:59:53 using the Galaxy Tool Factory.-->
  <description>Lifelines KM and optional Cox PH models</description>
  <requirements>
    <requirement version="1.5.3" type="package">pandas</requirement>
    <requirement version="3.7.2" type="package">matplotlib</requirement>
    <requirement version="0.27.7" type="package">lifelines</requirement>

    if colsok:
        df.columns = testcols # use actual header
    else:
        colsok = (args.time in defaultcols) and (args.status in defaultcols)
        if colsok:
            print('Replacing first row of data derived header %s with %s' % (testcols, defaultcols))
            df.columns = defaultcols
        else:
            sys.stderr.write('## CRITICAL USAGE ERROR (not a bug!): time %s and status %s do not match anything in the file header, supplied header or automatic default column names %s' % (args.time, args.status, defaultcols))
print('## Lifelines tool\nInput data header =', df.columns, 'time column =', args.time, 'status column =', args.status)
os.makedirs(args.image_dir, exist_ok=True)
fig, ax = plt.subplots()
if args.group > '':
    names = []
    times = []

    kmf.fit(df[args.time], df[args.status])
    kmf.plot_survival_function(ax=ax)
    ax.set_title(args.title)
    fig.savefig(os.path.join(args.image_dir,'KM_%s.png' % args.title))
    print('#### No grouping variable, so no log rank or other Kaplan-Meier statistical output is available')
survdf = lifelines.utils.survival_table_from_events(df[args.time], df[args.status])
lifedf = lifelines.utils.survival_table_from_events(df[args.time], df[args.status], collapse=True)
print("#### Survival table using time %s and event %s" % (args.time, args.status))
with pd.option_context('display.max_rows', None,
                       'display.max_columns', None,
                       'display.precision', 3,
                       ):
    print(survdf)
print("#### Life table using time %s and event %s" % (args.time, args.status))
with pd.option_context('display.max_rows', None,
                       'display.max_columns', None,
                       'display.precision', 3,
                       ):
    print(lifedf)
outpath = os.path.join(args.image_dir,'survival_table.tabular')
survdf.to_csv(outpath, sep='\t')
outpath = os.path.join(args.image_dir,'life_table.tabular')
lifedf.to_csv(outpath, sep='\t')
if len(args.cphcols) > 0:
    fig, ax = plt.subplots()
    ax.set_title('Cox-PH model: %s' % args.title)
    cphcols = args.cphcols.strip().split(',')
    cphcols = [x.strip() for x in cphcols]

    cphdf = df[cutcphcols]
    ucolcounts = colsdf.nunique(axis=0)
    cph.fit(cphdf, duration_col=args.time, event_col=args.status)
    cph.print_summary()
    for i, cov in enumerate(colsdf.columns):
         if ucolcounts[i] > 10: # a hack - assume categories are sparse - if not imaginary quintiles will have to do
             v = pd.Series.tolist(cphdf[cov].quantile(QVALS))
             vdt = df.dtypes[cov]
             if vdt == 'int64':
                 v = trimlegend(v)
             axp = cph.plot_partial_effects_on_outcome(cov, cmap='coolwarm', values=v)