Skip to content

Benchmark

benchmark(config, benchmark_type, output_dir, no_curves)

Benchmark results for specified runs for a specified prioritisation type for comparison. Args: config (Config): Configuration for benchmarking. benchmark_type (BenchmarkOutputType): Benchmark output type. output_dir (Path): Output directory for benchmarking results. no_curves (bool): Whether to skip generating binary classification curves.

Source code in src/pheval/analyse/benchmark.py 
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
def benchmark(config: Config, benchmark_type: BenchmarkOutputType, output_dir: Path, no_curves: bool) -> None:
    """
    Benchmark results for specified runs for a specified prioritisation type for comparison.
    Args:
        config (Config): Configuration for benchmarking.
        benchmark_type (BenchmarkOutputType): Benchmark output type.
        output_dir (Path): Output directory for benchmarking results.
        no_curves (bool): Whether to skip generating binary classification curves.
    """
    conn = duckdb.connect(output_dir.joinpath(f"{config.benchmark_name}.duckdb"))
    stats, curve_results, true_positive_cases = process_stats(config.runs, benchmark_type, no_curves)
    write_table(conn, stats, f"{config.benchmark_name}_{benchmark_type.prioritisation_type_string}_summary")
    if not no_curves:
        write_table(
            conn,
            curve_results,
            f"{config.benchmark_name}_{benchmark_type.prioritisation_type_string}_binary_classification_curves",
        )
    run_identifiers = [run.run_identifier for run in config.runs]
    calculate_rank_changes(conn, run_identifiers, true_positive_cases, benchmark_type)
    generate_plots(
        benchmark_name=config.benchmark_name,
        benchmarking_results_df=stats,
        curves=curve_results,
        benchmark_output_type=benchmark_type,
        plot_customisation=config.plot_customisation,
        output_dir=output_dir,
        no_curves=no_curves,
        conn=conn,
        run_identifiers=run_identifiers,
    )
    conn.close()

benchmark_runs(benchmark_config_file, output_dir, no_curves)

Benchmark results for specified runs for comparison. Args: benchmark_config_file (Path): Path to benchmark config file. output_dir (Path): Output directory for benchmarking results. no_curves (bool): Whether to skip generating binary classification curves.

Source code in src/pheval/analyse/benchmark.py 
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
def benchmark_runs(benchmark_config_file: Path, output_dir: Path, no_curves: bool) -> None:
    """
    Benchmark results for specified runs for comparison.
    Args:
        benchmark_config_file (Path): Path to benchmark config file.
        output_dir (Path): Output directory for benchmarking results.
        no_curves (bool): Whether to skip generating binary classification curves.
    """
    output_dir.mkdir(parents=True, exist_ok=True)
    logger = get_logger()
    start_time = time.perf_counter()
    logger.info("Initiated benchmarking process.")
    config = parse_run_config(benchmark_config_file)
    if Path(output_dir).joinpath(f"{config.benchmark_name}.duckdb").exists():
        logger.error(f"{config.benchmark_name}.duckdb already exists! Exiting.")
        sys.exit(1)
    gene_analysis_runs = [run for run in config.runs if run.gene_analysis]
    variant_analysis_runs = [run for run in config.runs if run.variant_analysis]
    disease_analysis_runs = [run for run in config.runs if run.disease_analysis]
    if gene_analysis_runs:
        logger.info("Initiating benchmarking for gene results.")
        benchmark(
            Config(
                benchmark_name=config.benchmark_name,
                runs=gene_analysis_runs,
                plot_customisation=config.plot_customisation,
            ),
            BenchmarkOutputTypeEnum.GENE.value,
            output_dir,
            no_curves,
        )
        logger.info("Finished benchmarking for gene results.")
    if variant_analysis_runs:
        logger.info("Initiating benchmarking for variant results")
        benchmark(
            Config(
                benchmark_name=config.benchmark_name,
                runs=variant_analysis_runs,
                plot_customisation=config.plot_customisation,
            ),
            BenchmarkOutputTypeEnum.VARIANT.value,
            output_dir,
            no_curves,
        )
        logger.info("Finished benchmarking for variant results.")
    if disease_analysis_runs:
        logger.info("Initiating benchmarking for disease results")
        benchmark(
            Config(
                benchmark_name=config.benchmark_name,
                runs=disease_analysis_runs,
                plot_customisation=config.plot_customisation,
            ),
            BenchmarkOutputTypeEnum.DISEASE.value,
            output_dir,
            no_curves,
        )
        logger.info("Finished benchmarking for disease results.")
    logger.info(f"Finished benchmarking! Total time: {time.perf_counter() - start_time:.2f} seconds.")

process_stats(runs, benchmark_type, no_curves)

Processes stats outputs for specified runs to compare. Args: runs (List[RunConfig]): List of runs to benchmark. benchmark_type (BenchmarkOutputTypeEnum): Benchmark output type. no_curves (bool): Whether to skip generating binary classification curves. Returns: Tuple[pl.DataFrame, pl.DataFrame, pl.DataFrame, pl.DataFrame]: The stats for all runs.

Source code in src/pheval/analyse/benchmark.py 
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
def process_stats(
    runs: list[RunConfig], benchmark_type: BenchmarkOutputType, no_curves: bool
) -> tuple[pl.DataFrame, pl.DataFrame, pl.DataFrame]:
    """
    Processes stats outputs for specified runs to compare.
    Args:
        runs (List[RunConfig]): List of runs to benchmark.
        benchmark_type (BenchmarkOutputTypeEnum): Benchmark output type.
        no_curves (bool): Whether to skip generating binary classification curves.
    Returns:
        Tuple[pl.DataFrame, pl.DataFrame, pl.DataFrame, pl.DataFrame]: The stats for all runs.
    """
    stats, curve_results, true_positive_cases = [], [], []
    for run in runs:
        result_scan = scan_directory(run, benchmark_type)
        stats.append(
            compute_rank_stats(run.run_identifier, result_scan).join(
                compute_confusion_matrix(run.run_identifier, result_scan), on="run_identifier"
            )
        )
        if not no_curves:
            curve_results.append(compute_curves(run.run_identifier, result_scan))
        true_positive_cases.append(
            result_scan.filter(pl.col("true_positive"))
            .select(["result_file", *benchmark_type.columns, pl.col("rank").alias(run.run_identifier)])
            .sort(["result_file", *benchmark_type.columns])
        )
    return (
        pl.concat(stats, how="vertical").collect(),
        pl.concat(curve_results, how="vertical").collect() if not no_curves else None,
        pl.concat(
            [true_positive_cases[0]]
            + [
                df.select(
                    [col for col in df.collect_schema().keys() if col not in ["result_file", *benchmark_type.columns]]
                )
                for df in true_positive_cases[1:]
            ],
            how="horizontal",
        ).collect(),
    )

scan_directory(run, benchmark_type)

Scan a results directory containing pheval parquet standardised results and return a LazyFrame object. Args: run (RunConfig): RunConfig object. benchmark_type (BenchmarkOutputTypeEnum): Benchmark output type. Returns: pl.LazyFrame: LazyFrame object containing all the results in the directory.

Source code in src/pheval/analyse/benchmark.py 
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
def scan_directory(run: RunConfig, benchmark_type: BenchmarkOutputType) -> pl.LazyFrame:
    """
    Scan a results directory containing pheval parquet standardised results and return a LazyFrame object.
    Args:
        run (RunConfig): RunConfig object.
        benchmark_type (BenchmarkOutputTypeEnum): Benchmark output type.
    Returns:
        pl.LazyFrame: LazyFrame object containing all the results in the directory.
    """
    logger = get_logger()
    logger.info(f"Analysing results in {run.results_dir.joinpath(benchmark_type.result_directory)}")

    lf = pl.scan_parquet(
        run.results_dir.joinpath(benchmark_type.result_directory),
        include_file_paths="file_path",
    ).with_columns(
        pl.col("rank").cast(pl.Int64),
        pl.col("file_path").str.extract(r"([^/\\]+)$").alias("result_file"),
        pl.col("true_positive").fill_null(False),
    )

    if run.threshold is None:
        passes_threshold = pl.lit(True)
    else:
        passes_threshold = (
            pl.col("score") >= pl.lit(run.threshold)
            if run.score_order.lower() == "descending"
            else pl.col("score") <= pl.lit(run.threshold)
        )

    lf = lf.filter(pl.col("true_positive") | passes_threshold)

    return (
        lf.filter(pl.col("true_positive") | passes_threshold)
        .with_columns(
            pl.when(pl.col("true_positive") & ~passes_threshold).then(pl.lit(0)).otherwise(pl.col("rank")).alias("rank")
        )
        .sort(by="score", descending=(run.score_order.lower() == "descending"))
        .unique(subset=_get_unique_subset(benchmark_type), keep="first")
    )