The following code I have developed with PYSAM looks for reads completely soft clipped

 import pysam
import re
import sys
import os

if len(sys.argv) != 2:
    print("Usage: python improve_bam.py input_bam_path")
    sys.exit(1)

# Input BAM file path provided as a command-line argument
input_bam_path = sys.argv[1]

# Generate the output BAM file path
output_directory = os.path.dirname(input_bam_path)
input_filename = os.path.basename(input_bam_path)
output_filename = input_filename.replace(".bam", "_improved.bam")
output_bam_path = os.path.join(output_directory, output_filename)

# Additional parameters
FLAG = 4  # Set FLAG to indicate an unmapped read
RNAME = -1  # Set RNAME to -1 to indicate unmapped
MAPQ = 0
CIGAR = "*"  # Set CIGAR to '*' to indicate no cigar string
RNEXT = -1  # Set RNEXT to -1 to indicate no "next read"
PNEXT = -1  # Set PNEXT to -1 to indicate no "next position"
TLEN = 0  # Set TLEN to 0 to indicate no template length
#POS = 0

print(f"Flag: {FLAG}")
print(f"MAPQ: {MAPQ}")
print(f"RNAME: {RNAME}")
print(f"RNEXT: {RNEXT}")
print(f"PNEXT: {PNEXT}")
print(f"TLEN: {TLEN}")

# Open input BAM file
input_bam = pysam.AlignmentFile(input_bam_path, "rb")

# Open output BAM file for writing
output_bam = pysam.AlignmentFile(output_bam_path, "wb", header=input_bam.header)

total_reads = 0
changed_reads = 0

for read in input_bam:
    total_reads += 1
    if not read.is_unmapped:
        # Check if the CIGAR string matches the pattern: <number>S
        cigar_string = read.cigarstring
        if cigar_string and re.match(r'^\d+S$', cigar_string):
            # Print the read before modification
            print("Before modification:")
            print(f"Flag: {read.flag}")
            print(f"MAPQ: {read.mapping_quality}")
            print(f"RNAME: {read.rname}")
            print(f"RNEXT: {read.rnext}")
            print(f"PNEXT: {read.pnext}")
            print(f"TLEN: {read.tlen}")
            print(f"POS: {read.pos}")

            # Perform the necessary modifications
            read.flag = FLAG
            read.mapping_quality = MAPQ
            read.rname = RNAME  
            read.rnext = RNEXT
            read.pnext = PNEXT
            read.tlen = TLEN
            read.cigarstring = CIGAR
            # read.pos = POS
            if read.rname == -1:
                read.is_unmapped = True
                read.pos = 0

            # Print the read after modification
            print("After modification:")
            print(f"Flag: {read.flag}")
            print(f"MAPQ: {read.mapping_quality}")
            print(f"RNAME: {read.rname}")
            print(f"RNEXT: {read.rnext}")
            print(f"PNEXT: {read.pnext}")
            print(f"TLEN: {read.tlen}")
            print(f"POS: {read.pos}")

            changed_reads += 1
    # Write the modified or unmodified read to the output BAM file
    output_bam.write(read)


# Close the input and output BAM files
input_bam.close()
output_bam.close()

# Print the total and changed reads counts
print("Total reads in the BAM file:", total_reads)
print("Changed reads:", changed_reads)

This seems to work and a normal output is like this

 Before modification:
Flag: 0
MAPQ: 0
RNAME: 16
RNEXT: -1
PNEXT: -1
TLEN: 0
POS: 27116631
After modification:
Flag: 4
MAPQ: 0
RNAME: -1
RNEXT: -1
PNEXT: -1
TLEN: 0
POS: 0

Apart from a new BAM file (called _improved.bam)

HOWEVER, (and this is what I don't understand) when I do a validation in the _improved.bam file I get this error

 ERROR::INVALID_ALIGNMENT_START:Record 1699229, Read name ML-PT1-10:12:000H00CHL:1:13107:9070:8182, Alignment start should be 0 because reference name = *.

If I convert the 2 BAM files to a SAM format and I look at that specific read I get this

BEFORE change:

 ML-PT1-10:12:000H00CHL:1:13107:9070:8182        256     chr17   7669612 0       151S    *       0       0       CCTAAGAGCAATCAGTGGGGAACAAGAAGTGGAGAATGTCATGGAATTCTCGGGTGCCAAGGAACTCCAGTCACTTAGGCATCTCGTATGCCGTCTTCTGCTTGAAAAAAAAAAGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGG     AAFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFA00AFFF0FFFFFFFFFAFFFFFFFFFFFFFF000FFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFF       SA:Z:chr2,32916401,+,113S38M,0,0;       RG:Z:MiniSeq_TruSight_Tumor_15_FFPE_and_control_samples_-27687666_HD729_S5andS6_L001_R1_001   UQ:i:74 AS:i:30

After the change:

 ML-PT1-10:12:000H00CHL:1:13107:9070:8182        4       *       1       0       *       *       0       0       CCTAAGAGCAATCAGTGGGGAACAAGAAGTGGAGAATGTCATGGAATTCTCGGGTGCCAAGGAACTCCAGTCACTTAGGCATCTCGTATGCCGTCTTCTGCTTGAAAAAAAAAAGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGG     AAFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFA00AFFF0FFFFFFFFFAFFFFFFFFFFFFFF000FFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFF       SA:Z:chr2,32916401,+,113S38M,0,0;       RG:Z:MiniSeq_TruSight_Tumor_15_FFPE_and_control_samples_-27687666_HD729_S5andS6_L001_R1_001   UQ:i:74 AS:i:30

That "1" in the fourth field should be "0" shouldn't be?? Why this has not been modified? Why did PYSAM report to me that this was 0 when it is not?

To give some context, what I am looking for is to NULL these reads to PASS validation and also to not get an error when this BAM is used by PISCES.