I am trying to run the Picard CollectMultipleMetrics command on a CRAM file to obtain alignment metrics at a read group level, using the accumulation level flag and the alignment metrics program:

-METRIC_ACCUMULATION_LEVEL READ_GROUP
-PROGRAM CollectAlignmentSummaryMetrics

The CRAM was aligned to the hg38 reference from multiple read groups. The different read groups can be seen in the CRAM header:

samtools view -H sample123.cram 
... 
@RG ID:sample123 LB:LB0 PL:PL0 PU:PU0 SM:sample123 
@RG ID:sample123-28923D42 LB:LB0 PL:PL0 PU:PU0 SM:sample123 
@RG ID:sample123-3CEDF6CF LB:LB0 PL:PL0 PU:PU0 SM:sample123 
@RG ID:sample123-5885F29A LB:LB0 PL:PL0 PU:PU0 SM:sample123 
...

I was expecting that the metrics would be accumulated for each RG ID present in the header.

However, it seems that the metrics are accumulated based on the PU (Platform Unit?) of each read group. Since each group has the same PU - "PU0" - the metrics for all read groups are being collapsed together and compressed into just one set of metrics, for "PU0" read groups. 

See below for the output. To me, it looks like the "PU" field is being used as the "READ_GROUP" ID.

Should I expect to see alignment metrics for each RG ID in the header, or is it normal for the program to group reads based on the PU field? If I did want to get alignment metrics for each different read group that went into the CRAM, how can this be done?

## htsjdk.samtools.metrics.StringHeader
# CollectMultipleMetrics INPUT=/io/batch/4e6e78/sample123.cram ASSUME_SORTED=true OUTPUT=/io/batch/4e6e78/prefix METRIC_ACCUMULATION_LEVEL=[READ_GROUP] PROGRAM=[CollectAlignmentSummaryMetrics, CollectInsertSizeMetrics, MeanQualityByCycle, CollectBaseDistributionByCycle, CollectQualityYieldMetrics] VALIDATION_STRINGENCY=SILENT REFERENCE_SEQUENCE=/io/batch/4e6e78/inputs/0v8nr/Homo_sapiens_assembly38_masked.fasta    STOP_AFTER=0 INCLUDE_UNPAIRED=false VERBOSITY=INFO QUIET=false COMPRESSION_LEVEL=5 MAX_RECORDS_IN_RAM=500000 CREATE_INDEX=false CREATE_MD5_FILE=false USE_JDK_DEFLATER=false USE_JDK_INFLATER=false
## htsjdk.samtools.metrics.StringHeader
# Started on: Fri Sep 20 02:06:35 UTC 2024

## METRICS CLASS    picard.analysis.AlignmentSummaryMetrics
CATEGORY    TOTAL_READS    PF_READS    PCT_PF_READS    PF_NOISE_READS    PF_READS_ALIGNED    PCT_PF_READS_ALIGNED    PF_ALIGNED_BASES    PF_HQ_ALIGNED_READS    PF_HQ_ALIGNED_BASES    PF_HQ_ALIGNED_Q20_BASES    PF_HQ_MEDIAN_MISMATCHES    PF_MISMATCH_RATE    PF_HQ_ERROR_RATE    PF_INDEL_RATE    MEAN_READ_LENGTH    SD_READ_LENGTH    MEDIAN_READ_LENGTH    MAD_READ_LENGTH    MIN_READ_LENGTH    MAX_READ_LENGTH    MEAN_ALIGNED_READ_LENGTH    READS_ALIGNED_IN_PAIRS    PCT_READS_ALIGNED_IN_PAIRS    PF_READS_IMPROPER_PAIRS    PCT_PF_READS_IMPROPER_PAIRS    BAD_CYCLES    STRAND_BALANCE    PCT_CHIMERAS    PCT_ADAPTER    PCT_SOFTCLIP    PCT_HARDCLIP    AVG_POS_3PRIME_SOFTCLIP_LENGTH    SAMPLE    LIBRARY    READ_GROUP
FIRST_OF_PAIR    244114554    244114554    1    0    243578766    0.997805    35830086505    223433790    32942266262    32484341761    0    0.003247    0.002252    0.00024    148    0    148    0    148    148    146.822344    242498638    0.995566    5273335    0.021649    0    0.50016    0.018886    0.000021    0.005762    0    36.276036    sample123    LB0    PU0
SECOND_OF_PAIR    244114554    244114554    1    0    242514944    0.993447    35411909816    221476505    32535845956    31197346891    0    0.005488    0.00445    0.00023    148    0    148    0    148    148    145.106691    242498638    0.999933    4209513    0.017358    0    0.499727    0.019563    0.000001    0.012997    0    40.307926    sample123    LB0    PU0
PAIR    488229108    488229108    1    0    486093710    0.995626    71241996321    444910295    65478112218    63681688652    0    0.004361    0.003344    0.000235    148    0    148    0    148    148    145.964517    484997276    0.997744    9482848    0.019508    0    0.499944    0.019224    0.000011    0.009379    0    38.920755    sample123    LB0    PU0

REQUIRED for all errors and issues:
a) GATK version used: 4.6.0 / Picard 3.2.0
b) Exact command used: 

CollectMultipleMetrics -INPUT /io/sample123.cram -REFERENCE_SEQUENCE /io/Homo_sapiens_assembly38_masked.fasta -OUTPUT /io/prefix -ASSUME_SORTED True -PROGRAM null -VALIDATION_STRINGENCY SILENT -PROGRAM CollectAlignmentSummaryMetrics -PROGRAM CollectInsertSizeMetrics -PROGRAM MeanQualityByCycle -PROGRAM CollectBaseDistributionByCycle -PROGRAM CollectQualityYieldMetrics -METRIC_ACCUMULATION_LEVEL null -METRIC_ACCUMULATION_LEVEL READ_GROUP

c) Entire program log (Filled with ProcessExecutor warnings but I'm not sure they're relevant)

INFO    2024-09-20 02:06:35    CollectMultipleMetrics    

INFO    2024-09-20 02:06:35    RExecutor    Executing R script via command: Rscript /tmp/script16343948897468350944.R
ERROR    2024-09-20 02:06:35    ProcessExecutor    During startup - Warning messages:
ERROR    2024-09-20 02:06:35    ProcessExecutor    1: Setting LC_CTYPE failed, using "C" 
ERROR    2024-09-20 02:06:35    ProcessExecutor    2: Setting LC_COLLATE failed, using "C" 
ERROR    2024-09-20 02:06:35    ProcessExecutor    3: Setting LC_TIME failed, using "C" 
ERROR    2024-09-20 02:06:35    ProcessExecutor    4: Setting LC_MESSAGES failed, using "C" 
ERROR    2024-09-20 02:06:35    ProcessExecutor    5: Setting LC_MONETARY failed, using "C" 
ERROR    2024-09-20 02:06:35    ProcessExecutor    6: Setting LC_PAPER failed, using "C" 
ERROR    2024-09-20 02:06:35    ProcessExecutor    7: Setting LC_MEASUREMENT failed, using "C" 
INFO    2024-09-20 02:06:35    ProcessExecutor    [1] "Checking if R is installed"
02:06:35.645 INFO  NativeLibraryLoader - Loading libgkl_compression.so from jar:file:/root/micromamba/share/picard-3.2.0-0/picard.jar!/com/intel/gkl/native/libgkl_compression.so
[Fri Sep 20 02:06:35 UTC 2024] CollectMultipleMetrics INPUT=/io/batch/4e6e78/sample123.cram ASSUME_SORTED=true OUTPUT=/io/batch/4e6e78/prefix METRIC_ACCUMULATION_LEVEL=[READ_GROUP] PROGRAM=[CollectAlignmentSummaryMetrics, CollectInsertSizeMetrics, MeanQualityByCycle, CollectBaseDistributionByCycle, CollectQualityYieldMetrics] VALIDATION_STRINGENCY=SILENT REFERENCE_SEQUENCE=/io/batch/4e6e78/inputs/0v8nr/Homo_sapiens_assembly38_masked.fasta    STOP_AFTER=0 INCLUDE_UNPAIRED=false VERBOSITY=INFO QUIET=false COMPRESSION_LEVEL=5 MAX_RECORDS_IN_RAM=500000 CREATE_INDEX=false CREATE_MD5_FILE=false USE_JDK_DEFLATER=false USE_JDK_INFLATER=false
[Fri Sep 20 02:06:35 UTC 2024] Executing as root@hostname-ec03d77033 on Linux 6.5.0-1023-gcp amd64; OpenJDK 64-Bit Server VM 22.0.1-internal-adhoc.conda.src; Deflater: Intel; Inflater: Intel; Provider GCS is available; Picard version: 3.2.0-1-g3948afb6b
WARNING    2024-09-20 02:06:35    CollectMultipleMetrics    The MeanQualityByCycle program does not support a metric accumulation level, but METRIC_ACCUMULATION_LEVEL was overridden in the command line. MeanQualityByCycle will be run against the entire input.
WARNING    2024-09-20 02:06:35    CollectMultipleMetrics    The CollectBaseDistributionByCycle program does not support a metric accumulation level, but METRIC_ACCUMULATION_LEVEL was overridden in the command line. CollectBaseDistributionByCycle will be run against the entire input.
WARNING    2024-09-20 02:06:35    CollectMultipleMetrics    The CollectQualityYieldMetrics program does not support a metric accumulation level, but METRIC_ACCUMULATION_LEVEL was overridden in the command line. CollectQualityYieldMetrics will be run against the entire input.
ERROR    2024-09-20 02:06:36    CollectAlignmentSummaryMetrics    ReadLength histogram is calculated on all reads only, but ALL_READS were not included in the Metric Accumulation Levels. Histogram will not be generated.
INFO    2024-09-20 02:07:02    SinglePassSamProgram    Processed     1,000,000 records.  Elapsed time: 00:00:25s.  Time for last 1,000,000:   22s.  Last read position: chr1:7,174,235
INFO    2024-09-20 02:07:23    SinglePassSamProgram    Processed     2,000,000 records.  Elapsed time: 00:00:46s.  Time for last 1,000,000:   20s.  Last read position: chr1:13,807,623
...
INFO    2024-09-20 03:57:47    SinglePassSamProgram    Processed   489,000,000 records.  Elapsed time: 01:51:10s.  Time for last 1,000,000:   19s.  Last read position: chrUn_JTFH01001929v1_decoy:599
INFO    2024-09-20 03:58:07    SinglePassSamProgram    Processed   490,000,000 records.  Elapsed time: 01:51:31s.  Time for last 1,000,000:   20s.  Last read position: */*
INFO    2024-09-20 03:59:01    RExecutor    Executing R script via command: Rscript /tmp/script9948235460342383020.R /io/batch/4e6e78/prefix.insert_size_metrics /io/batch/4e6e78/prefix.insert_size_histogram.pdf sample123.cram
ERROR    2024-09-20 03:59:01    ProcessExecutor    During startup - Warning messages:
ERROR    2024-09-20 03:59:01    ProcessExecutor    1: Setting LC_CTYPE failed, using "C" 
ERROR    2024-09-20 03:59:01    ProcessExecutor    2: Setting LC_COLLATE failed, using "C" 
ERROR    2024-09-20 03:59:01    ProcessExecutor    3: Setting LC_TIME failed, using "C" 
ERROR    2024-09-20 03:59:01    ProcessExecutor    4: Setting LC_MESSAGES failed, using "C" 
ERROR    2024-09-20 03:59:01    ProcessExecutor    5: Setting LC_MONETARY failed, using "C" 
ERROR    2024-09-20 03:59:01    ProcessExecutor    6: Setting LC_PAPER failed, using "C" 
ERROR    2024-09-20 03:59:01    ProcessExecutor    7: Setting LC_MEASUREMENT failed, using "C" 
INFO    2024-09-20 03:59:01    ProcessExecutor    null device 
INFO    2024-09-20 03:59:01    ProcessExecutor              1 
INFO    2024-09-20 03:59:01    RExecutor    Executing R script via command: Rscript /tmp/script1233926558258630103.R /io/batch/4e6e78/prefix.quality_by_cycle_metrics /io/batch/4e6e78/prefix.quality_by_cycle.pdf sample123.cram 
ERROR    2024-09-20 03:59:01    ProcessExecutor    During startup - Warning messages:
ERROR    2024-09-20 03:59:01    ProcessExecutor    1: Setting LC_CTYPE failed, using "C" 
ERROR    2024-09-20 03:59:01    ProcessExecutor    2: Setting LC_COLLATE failed, using "C" 
ERROR    2024-09-20 03:59:01    ProcessExecutor    3: Setting LC_TIME failed, using "C" 
ERROR    2024-09-20 03:59:01    ProcessExecutor    4: Setting LC_MESSAGES failed, using "C" 
ERROR    2024-09-20 03:59:01    ProcessExecutor    5: Setting LC_MONETARY failed, using "C" 
ERROR    2024-09-20 03:59:01    ProcessExecutor    6: Setting LC_PAPER failed, using "C" 
ERROR    2024-09-20 03:59:01    ProcessExecutor    7: Setting LC_MEASUREMENT failed, using "C" 
INFO    2024-09-20 03:59:01    ProcessExecutor    null device 
INFO    2024-09-20 03:59:01    ProcessExecutor              1 
INFO    2024-09-20 03:59:02    RExecutor    Executing R script via command: Rscript /tmp/script16334611869075646450.R /io/batch/4e6e78/prefix.base_distribution_by_cycle_metrics /io/batch/4e6e78/prefix.base_distribution_by_cycle.pdf sample123.cram 
ERROR    2024-09-20 03:59:02    ProcessExecutor    During startup - Warning messages:
ERROR    2024-09-20 03:59:02    ProcessExecutor    1: Setting LC_CTYPE failed, using "C" 
ERROR    2024-09-20 03:59:02    ProcessExecutor    2: Setting LC_COLLATE failed, using "C" 
ERROR    2024-09-20 03:59:02    ProcessExecutor    3: Setting LC_TIME failed, using "C" 
ERROR    2024-09-20 03:59:02    ProcessExecutor    4: Setting LC_MESSAGES failed, using "C" 
ERROR    2024-09-20 03:59:02    ProcessExecutor    5: Setting LC_MONETARY failed, using "C" 
ERROR    2024-09-20 03:59:02    ProcessExecutor    6: Setting LC_PAPER failed, using "C" 
ERROR    2024-09-20 03:59:02    ProcessExecutor    7: Setting LC_MEASUREMENT failed, using "C" 
INFO    2024-09-20 03:59:02    ProcessExecutor    null device 
INFO    2024-09-20 03:59:02    ProcessExecutor              1 
[Fri Sep 20 03:59:02 UTC 2024] picard.analysis.CollectMultipleMetrics done. Elapsed time: 112.45 minutes.
Runtime.totalMemory()=6501171200