Lab BDS Build

1library(yamaa)23adlb <- yamaa_domain("spec.yaml")$output4adlb

1import yamaa23adlb = yamaa.yamaa_domain("spec.yaml").output4adlb

1# Reference solution for the yamaa benchmark adam-adlb-bds (R track).2#3# Written from the benchmark's full prompt and its inputs alone, as an agent4# would write it. Harbor's oracle agent runs it to check that the prompt can5# be solved and that the grader scores a correct answer 1.67library(dplyr, warn.conflicts = FALSE)8library(readr)910adsl <- read_csv(11  "/app/input/adsl.csv",12  col_types = cols(TRTSDT = col_date(), .default = col_character())13)14lb <- read_csv(15  "/app/input/lb.csv",16  col_types = cols(17    LBDTC = col_date(),18    LBSTRESN = col_double(),19    .default = col_character()20  )21)2223# One record per collected ALT/AST result; a collected result with no24# numeric value produces no record.25collected <- lb |>26  filter(LBTESTCD %in% c("ALT", "AST"), !is.na(LBSTRESN))2728# ALT and AST hold the collected result and unit; ALTSI holds the ALT29# result times 0.0167 with unit ukat/L, one record for each ALT record.30alt <- collected |>31  filter(LBTESTCD %in% "ALT") |>32  transmute(33    STUDYID, USUBJID,34    PARAMCD = "ALT",35    PARAM = "Alanine Aminotransferase",36    ADT = LBDTC,37    AVAL = LBSTRESN,38    AVALU = LBSTRESU39  )40ast <- collected |>41  filter(LBTESTCD %in% "AST") |>42  transmute(43    STUDYID, USUBJID,44    PARAMCD = "AST",45    PARAM = "Aspartate Aminotransferase",46    ADT = LBDTC,47    AVAL = LBSTRESN,48    AVALU = LBSTRESU49  )50altsi <- collected |>51  filter(LBTESTCD %in% "ALT") |>52  transmute(53    STUDYID, USUBJID,54    PARAMCD = "ALTSI",55    PARAM = "Alanine Aminotransferase (SI)",56    ADT = LBDTC,57    AVAL = LBSTRESN * 0.0167,58    AVALU = "ukat/L"59  )6061base <- bind_rows(alt, ast, altsi) |>62  left_join(63    adsl |> select(STUDYID, USUBJID, TRTSDT, TRT01A),64    by = c("STUDYID", "USUBJID")65  )6667# ABLFL is Y on the latest record on or before the treatment start date68# for each subject and parameter, including the start date itself.69baseline_dates <- base |>70  filter(ADT <= TRTSDT) |>71  group_by(STUDYID, USUBJID, PARAMCD) |>72  summarise(BASE_DT = max(ADT), .groups = "drop")7374with_flag <- base |>75  left_join(baseline_dates, by = c("STUDYID", "USUBJID", "PARAMCD")) |>76  mutate(ABLFL = if_else(!is.na(BASE_DT) & ADT == BASE_DT, "Y", NA_character_))7778# BASE repeats the flagged baseline value; CHG and PCHG follow, with PCHG79# empty when the baseline is zero. Values are not rounded.80baselines <- with_flag |>81  filter(ABLFL %in% "Y") |>82  select(STUDYID, USUBJID, PARAMCD, BASE = AVAL)8384adlb_flagged <- with_flag |>85  select(-BASE_DT) |>86  left_join(baselines, by = c("STUDYID", "USUBJID", "PARAMCD")) |>87  mutate(88    CHG = AVAL - BASE,89    PCHG = if_else(90      is.na(BASE) | BASE == 0,91      NA_real_,92      100 * (AVAL - BASE) / BASE93    )94  )9596# ASEQ numbers each subject's records from 1, ordered by PARAMCD (ALT,97# then ALTSI, then AST) and then by ADT.98param_order <- c("ALT" = 0L, "ALTSI" = 1L, "AST" = 2L)99100adlb <- adlb_flagged |>101  mutate(PARAM_ORDER = param_order[PARAMCD]) |>102  arrange(STUDYID, USUBJID, PARAM_ORDER, ADT) |>103  group_by(STUDYID, USUBJID) |>104  mutate(ASEQ = row_number()) |>105  ungroup() |>106  arrange(PARAM_ORDER, USUBJID, ADT) |>107  select(108    STUDYID, USUBJID, PARAMCD, PARAM, ADT, TRTSDT, TRT01A,109    AVAL, AVALU, ABLFL, BASE, CHG, PCHG, ASEQ110  )111112dir.create("/app/output", showWarnings = FALSE)113write_csv(adlb, "/app/output/adlb.csv", na = "")

1# Reference solution for the yamaa benchmark adam-adlb-bds (Python track).2#3# Written from the benchmark's full prompt and its inputs alone, as an agent4# would write it. Harbor's oracle agent runs it to check that the prompt can5# be solved and that the grader scores a correct answer 1.67from pathlib import Path89import polars as pl1011adsl = pl.read_csv("/app/input/adsl.csv", infer_schema=False).with_columns(12    pl.col("TRTSDT").str.to_date(),13)14lb = pl.read_csv("/app/input/lb.csv", infer_schema=False).with_columns(15    pl.col("LBSTRESN").cast(pl.Float64, strict=False).alias("RESULT"),16    pl.col("LBDTC").str.to_date().alias("ADT"),17)1819# One record per collected ALT/AST result; a collected result with no20# numeric value produces no record.21collected = lb.filter(22    pl.col("LBTESTCD").is_in(["ALT", "AST"]) & pl.col("RESULT").is_not_null()23)2425# ALT and AST hold the collected result and unit; ALTSI holds the ALT26# result times 0.0167 with unit ukat/L, one record for each ALT record.27alt = collected.filter(pl.col("LBTESTCD") == "ALT").with_columns(28    PARAMCD=pl.lit("ALT"),29    PARAM=pl.lit("Alanine Aminotransferase"),30    AVAL=pl.col("RESULT"),31    AVALU=pl.col("LBSTRESU"),32)33ast = collected.filter(pl.col("LBTESTCD") == "AST").with_columns(34    PARAMCD=pl.lit("AST"),35    PARAM=pl.lit("Aspartate Aminotransferase"),36    AVAL=pl.col("RESULT"),37    AVALU=pl.col("LBSTRESU"),38)39altsi = collected.filter(pl.col("LBTESTCD") == "ALT").with_columns(40    PARAMCD=pl.lit("ALTSI"),41    PARAM=pl.lit("Alanine Aminotransferase (SI)"),42    AVAL=pl.col("RESULT") * 0.0167,43    AVALU=pl.lit("ukat/L"),44)4546base = (47    pl.concat([alt, ast, altsi], how="diagonal")48    .select(49        "STUDYID",50        "USUBJID",51        "PARAMCD",52        "PARAM",53        "ADT",54        "AVAL",55        "AVALU",56    )57    .join(58        adsl.select("STUDYID", "USUBJID", "TRTSDT", "TRT01A"),59        on=["STUDYID", "USUBJID"],60        how="left",61    )62)6364# ABLFL is Y on the latest record on or before the treatment start date65# for each subject and parameter, including the start date itself.66baseline_dates = (67    base.filter(pl.col("ADT") <= pl.col("TRTSDT"))68    .group_by("STUDYID", "USUBJID", "PARAMCD")69    .agg(pl.col("ADT").max().alias("BASE_DT"))70)71with_flag = base.join(72    baseline_dates, on=["STUDYID", "USUBJID", "PARAMCD"], how="left"73).with_columns(74    ABLFL=pl.when(pl.col("ADT") == pl.col("BASE_DT"))75    .then(pl.lit("Y"))76    .otherwise(None)77)7879# BASE repeats the flagged baseline value; CHG and PCHG follow, with PCHG80# empty when the baseline is zero. Values are not rounded.81baselines = (82    with_flag.filter(pl.col("ABLFL") == "Y")83    .select("STUDYID", "USUBJID", "PARAMCD", BASE="AVAL")84)85adlb_flagged = (86    with_flag.drop("BASE_DT")87    .join(baselines, on=["STUDYID", "USUBJID", "PARAMCD"], how="left")88    .with_columns(89        CHG=pl.col("AVAL") - pl.col("BASE"),90        PCHG=pl.when(pl.col("BASE").is_null() | (pl.col("BASE") == 0))91        .then(None)92        .otherwise(100 * (pl.col("AVAL") - pl.col("BASE")) / pl.col("BASE")),93    )94)9596# ASEQ numbers each subject's records from 1, ordered by PARAMCD (ALT,97# then ALTSI, then AST) and then by ADT.98order = {"ALT": 0, "ALTSI": 1, "AST": 2}99adlb = (100    adlb_flagged.with_columns(101        PARAM_ORDER=pl.col("PARAMCD").replace_strict(order)102    )103    .sort(["STUDYID", "USUBJID", "PARAM_ORDER", "ADT"])104    .with_columns(105        ASEQ=pl.int_range(1, pl.len() + 1).over(["STUDYID", "USUBJID"]).cast(pl.Int64)106    )107    .sort(["PARAM_ORDER", "USUBJID", "ADT"])108    .select(109        "STUDYID",110        "USUBJID",111        "PARAMCD",112        "PARAM",113        "ADT",114        "TRTSDT",115        "TRT01A",116        "AVAL",117        "AVALU",118        "ABLFL",119        "BASE",120        "CHG",121        "PCHG",122        "ASEQ",123    )124)125126Path("/app/output").mkdir(exist_ok=True)127adlb.write_csv("/app/output/adlb.csv")

1/* opensas reference solution for the yamaa benchmark adam-adlb-bds.2   Reads the benchmark inputs and derives the requested output.3   Execute with the existing opensas runtime: opensas /app/output/result.sas. */45data adsl;6  length STUDYID USUBJID TRTSDT TRT01A $1024;7  infile "/app/input/adsl.csv" dsd dlm="," firstobs=2 truncover lrecl=32767;8  input STUDYID : $char1024. USUBJID : $char1024. TRTSDT : $char1024. TRT01A : $char1024.;9run;1011data lb;12  length STUDYID USUBJID LBTESTCD LBTEST LBDTC LBSTRESU $1024;13  infile "/app/input/lb.csv" dsd dlm="," firstobs=2 truncover lrecl=32767;14  input STUDYID : $char1024. USUBJID : $char1024. LBTESTCD : $char1024. LBTEST : $char1024. LBDTC : $char1024. LBSTRESN LBSTRESU : $char1024.;15run;1617data parameters;18  length PARAMCD $16 PARAM $40 ADT $10 AVALU $1024;19  set lb;20  if LBTESTCD in ('ALT','AST') and not missing(LBSTRESN);21  ADT=LBDTC; PARAMCD=LBTESTCD; AVAL=LBSTRESN; AVALU=LBSTRESU;22  if LBTESTCD='ALT' then PARAM='Alanine Aminotransferase'; else PARAM='Aspartate Aminotransferase';23  output;24  if LBTESTCD='ALT' then do; PARAMCD='ALTSI'; PARAM='Alanine Aminotransferase (SI)'; AVAL=LBSTRESN*0.0167; AVALU='ukat/L'; output; end;25run;26proc sql; create table joined as select a.*,b.TRTSDT,b.TRT01A from parameters as a left join adsl as b on a.STUDYID=b.STUDYID and a.USUBJID=b.USUBJID; quit;27proc sort data=joined(where=(ADT ne '' and TRTSDT ne '' and ADT<=TRTSDT)) out=candidates; by STUDYID USUBJID PARAMCD descending ADT; run;28data baseline; set candidates; by STUDYID USUBJID PARAMCD; if first.PARAMCD; BASE=AVAL; keep STUDYID USUBJID PARAMCD ADT BASE; rename ADT=BASEDT; run;29proc sql; create table combined as select a.*,b.BASE,b.BASEDT from joined as a left join baseline as b on a.STUDYID=b.STUDYID and a.USUBJID=b.USUBJID and a.PARAMCD=b.PARAMCD; quit;30proc sort data=combined; by STUDYID USUBJID PARAMCD ADT; run;31data result;32  length ABLFL $1;33  set combined; by STUDYID USUBJID;34  if first.USUBJID then ASEQ=0; ASEQ+1;35  if ADT=BASEDT and not missing(BASEDT) then ABLFL='Y';36  CHG=AVAL-BASE;37  if not missing(BASE) and BASE ne 0 then PCHG=100*CHG/BASE;38run;3940proc sql;41  create table final as select STUDYID, USUBJID, PARAMCD, PARAM, ADT, TRTSDT, TRT01A, AVAL, AVALU, ABLFL, BASE, CHG, PCHG, ASEQ from result;42quit;43proc export data=final outfile="/app/output/adlb.csv" dbms=csv replace; run;