extract_regions_bw ValueError: could not broadcast array (880→1000 / 740→800) — shape mismatch at chrom ends

[sbg15236]

Hi,
I’m getting a consistent shape mismatch when running finemo extract-regions-bw, even though all inputs (FASTA, .fai, bigWig, and narrowPeak) come from the same mm39 reference genome.

Error examples:
(Region width 1000) → ValueError: could not broadcast input array from shape (880,) into shape (1000,)
(Region width 800) → ValueError: could not broadcast input array from shape (740,) into shape (800,)

Checks performed

FASTA and bigWig use the same mm39 reference

Chrom sizes verified identical

All peaks confined within min(FASTA_len, BW_len)

Pre-validated with pyBigWig: every region returns the expected window size (800 or 1000)

Still fails inside Fi-NeMo

Environment:

Fi-NeMo GPU: kundajelab/finemo_gpu:latest

Runtime: Singularity / RHEL9 (A100 GPU)

Python: 3.10

bigWig: ChromBPNet projected contributions (mm39)

Please advise!

Best,

Steve

[austintwang]

Hi Steve, happy to help. Do you mind attaching the full traceback as well as a minimal peak region coordinate causing the error?

Thanks!

[sbg15236]

Thanks Austin! I scanned my peaks file to find an offending peak and found failure at this peak as en example, chrM 10992 16275 peak_calling_CD19cre_cluster_0_peak_129924bb 15835 . 2.21613 1583.59 1580.15 4927 The full traceback I have is as follows: Traceback (most recent call last): File "/opt/conda/envs/finemo/bin/finemo", line 8, in <module> sys.exit(cli()) File "/opt/finemo_gpu/src/finemo/main.py", line 349, in cli extract_regions_bw(args.peaks, args.fasta, args.bigwigs, args.out_path, args.region_width) File "/opt/finemo_gpu/src/finemo/main.py", line 11, in extract_regions_bw sequences, contribs = data_io.load_regions_from_bw(peaks_df, fa_path, bw_paths, half_width) File "/opt/finemo_gpu/src/finemo/data_io.py", line 109, in load_regions_from_bw contrib_buffer[j,:] = np.nan_to_num(bw.values(chrom, start_adj, end_adj, numpy=True)) ValueError: could not broadcast input array from shape (740,) into shape (800,)

…

On Tue, Nov 4, 2025 at 10:44 PM Austin Wang ***@***.***> wrote: *austintwang* left a comment (austintwang/finemo_gpu#42) <#42 (comment)> Hi Steve, happy to help. Do you mind attaching the full traceback as well as a minimal peak region coordinate causing the error? Thanks! — Reply to this email directly, view it on GitHub <#42 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AISOKIMOYEUX6WJG7ZHFDFD33FXA5AVCNFSM6AAAAACLEID6JOVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZTIOBZGEZTGNZUHA> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[austintwang]

Hi Steve, I implemented a fix in branch bw_chrom_edge. Could you confirm whether this addresses your error?

Thanks!

[austintwang]

Actually, before you do so, do you mind if you see the same issue here?
https://github.com/kundajelab/Fi-NeMo

This is a newer version that's more actively maintained

[sbg15236]

Hi Austin, Sorry for the delay, it's been busy here. Yes I will try running the newer version and see if the same problem occurs again. I'll try to do it today, but if not then early next week! Best, Steve

…

On Wed, Nov 5, 2025 at 9:09 PM Austin Wang ***@***.***> wrote: *austintwang* left a comment (austintwang/finemo_gpu#42) <#42 (comment)> Actually, before you do so, do you mind if you see the same issue here? https://github.com/kundajelab/Fi-NeMo This is a newer version that's more actively maintained — Reply to this email directly, view it on GitHub <#42 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AISOKIOC4P6WPKCDLX2N7VL33KUU3AVCNFSM6AAAAACLEID6JOVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZTIOJUGUZTCMJQG4> . You are receiving this because you authored the thread.Message ID: ***@***.***>