Geometric mean metabolite ratio and replicate variability per subject

1library(yamaa)23yamaa_spec("spec.yaml")$run()$output

1import yamaa23yamaa.spec("spec.yaml").run().output

1# Reference solution for the yamaa benchmark adam-adpp-geometric-mean-ratio (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)910pp <- read_csv(11  "/app/input/pp.csv",12  col_types = cols(PPSTRESN = col_double(), .default = col_character())13)14pc <- read_csv(15  "/app/input/pc.csv",16  col_types = cols(PCSTRESN = col_double(), .default = col_character())17)1819geomean <- pp |>20  group_by(STUDYID, USUBJID) |>21  summarise(22    AVAL = {23      x <- PPSTRESN[PPTESTCD == "MRCMAX" & !is.na(PPSTRESN) & PPSTRESN > 0]24      if (length(x) > 0) exp(mean(log(x))) else NA_real_25    },26    .groups = "drop"27  ) |>28  mutate(29    PARAMCD = "MRCMAXGM",30    PARAM = "Geometric Mean Metabolite Ratio Cmax",31    AVALU = "RATIO"32  )3334replicate_sd <- pc |>35  group_by(STUDYID, USUBJID) |>36  summarise(37    AVAL = {38      x <- PCSTRESN[PCTPT == "2H" & !is.na(PCSTRESN)]39      if (length(x) >= 2) sd(x) else NA_real_40    },41    .groups = "drop"42  ) |>43  mutate(44    PARAMCD = "C2HREPSD",45    PARAM = "Replicate SD of 2 H Concentration",46    AVALU = "ng/mL"47  )4849adpp <- bind_rows(geomean, replicate_sd) |>50  arrange(STUDYID, USUBJID, PARAMCD) |>51  select(STUDYID, USUBJID, PARAMCD, PARAM, AVAL, AVALU)5253dir.create("/app/output", showWarnings = FALSE)54write_csv(adpp, "/app/output/adpp.csv", na = "")

1# Reference solution for the yamaa benchmark adam-adpp-geometric-mean-ratio (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 1.67import csv8import math9import statistics1011PP_CSV = "/app/input/pp.csv"12PC_CSV = "/app/input/pc.csv"13OUTPUT_CSV = "/app/output/adpp.csv"141516def read(path):17    with open(path, newline="") as f:18        return list(csv.DictReader(f))192021def number(text):22    text = text.strip()23    return float(text) if text else None242526def cell(value):27    return "" if value is None else repr(value)282930def main():31    pp = read(PP_CSV)32    pc = read(PC_CSV)33    subjects = []34    for row in pp + pc:35        key = (row["STUDYID"], row["USUBJID"])36        if key not in subjects:37            subjects.append(key)3839    out_rows = []40    for study, subject in subjects:41        ratios = [42            number(r["PPSTRESN"])43            for r in pp44            if (r["STUDYID"], r["USUBJID"]) == (study, subject)45            and r["PPTESTCD"] == "MRCMAX"46        ]47        ratios = [x for x in ratios if x is not None and x > 0]48        geomean = (49            math.exp(sum(math.log(x) for x in ratios) / len(ratios)) if ratios else None50        )51        out_rows.append(52            {53                "STUDYID": study,54                "USUBJID": subject,55                "PARAMCD": "MRCMAXGM",56                "PARAM": "Geometric Mean Metabolite Ratio Cmax",57                "AVAL": cell(geomean),58                "AVALU": "RATIO",59            }60        )61        replicates = [62            number(r["PCSTRESN"])63            for r in pc64            if (r["STUDYID"], r["USUBJID"]) == (study, subject) and r["PCTPT"] == "2H"65        ]66        replicates = [x for x in replicates if x is not None]67        sd = statistics.stdev(replicates) if len(replicates) >= 2 else None68        out_rows.append(69            {70                "STUDYID": study,71                "USUBJID": subject,72                "PARAMCD": "C2HREPSD",73                "PARAM": "Replicate SD of 2 H Concentration",74                "AVAL": cell(sd),75                "AVALU": "ng/mL",76            }77        )7879    fieldnames = ["STUDYID", "USUBJID", "PARAMCD", "PARAM", "AVAL", "AVALU"]80    with open(OUTPUT_CSV, "w", newline="") as f:81        writer = csv.DictWriter(f, fieldnames=fieldnames)82        writer.writeheader()83        writer.writerows(out_rows)848586if __name__ == "__main__":87    main()

1/* opensas reference solution for the yamaa benchmark adam-adpp-geometric-mean-ratio.2   Reads the benchmark inputs and derives the requested output.3   Execute with the existing opensas runtime: opensas /app/output/result.sas. */45data pp;6  length STUDYID DOMAIN USUBJID PPTESTCD PPTEST PPSTRESU PPSTAT VISIT $40;7  infile "/app/input/pp.csv" dsd dlm="," firstobs=2 truncover lrecl=32767;8  input STUDYID : $char40. DOMAIN : $char40. USUBJID : $char40. PPSEQ9        PPTESTCD : $char40. PPTEST : $char40. PPSTRESN PPSTRESU : $char40.10        PPSTAT : $char40. VISITNUM VISIT : $char40.;11run;1213data pc;14  length STUDYID DOMAIN USUBJID PCTESTCD PCTEST PCSTRESU VISIT PCTPT $40;15  infile "/app/input/pc.csv" dsd dlm="," firstobs=2 truncover lrecl=32767;16  input STUDYID : $char40. DOMAIN : $char40. USUBJID : $char40. PCSEQ17        PCTESTCD : $char40. PCTEST : $char40. PCSTRESN PCSTRESU : $char40.18        PCREPNUM VISITNUM VISIT : $char40. PCTPT : $char40. PCTPTNUM;19run;2021/* The natural logarithm of each metabolite ratio above zero. */22data logs;23  set pp;24  if PPTESTCD = 'MRCMAX' and not missing(PPSTRESN) and PPSTRESN > 0;25  LOGRATIO = log(PPSTRESN);26run;2728/* The replicate 2-hour concentrations and their mean per subject. */29data reps;30  set pc;31  if PCTPT = '2H' and not missing(PCSTRESN);32run;3334proc sql;35  create table means as36    select STUDYID, USUBJID, mean(PCSTRESN) as M, count(*) as N37    from reps group by STUDYID, USUBJID;38  create table squares as39    select r.STUDYID, r.USUBJID, m.N,40           (r.PCSTRESN - m.M) * (r.PCSTRESN - m.M) as SQ41    from reps as r inner join means as m42      on m.STUDYID = r.STUDYID and m.USUBJID = r.USUBJID;43  create table subjects as44    select distinct STUDYID, USUBJID from pp45    union46    select distinct STUDYID, USUBJID from pc;47  create table geomean as48    select STUDYID, USUBJID, exp(mean(LOGRATIO)) as AVAL49    from logs group by STUDYID, USUBJID;50  create table sd as51    select STUDYID, USUBJID,52           case when N >= 2 then sqrt(sum(SQ) / (N - 1)) else . end as AVAL53    from squares group by STUDYID, USUBJID, N;54  create table final as55    select s.STUDYID, s.USUBJID, 'MRCMAXGM' as PARAMCD length=8,56           'Geometric Mean Metabolite Ratio Cmax' as PARAM length=40,57           g.AVAL, 'RATIO' as AVALU length=858    from subjects as s left join geomean as g59      on g.STUDYID = s.STUDYID and g.USUBJID = s.USUBJID60    union all61    select s.STUDYID, s.USUBJID, 'C2HREPSD' as PARAMCD length=8,62           'Replicate SD of 2 H Concentration' as PARAM length=40,63           d.AVAL, 'ng/mL' as AVALU length=864    from subjects as s left join sd as d65      on d.STUDYID = s.STUDYID and d.USUBJID = s.USUBJID66    order by 1, 2, 3;67quit;6869proc export data=final outfile="/app/output/adpp.csv" dbms=csv replace; run;