error on 'Normalizing raw contacts by HiCzin'

[Taojianchang]

Dear sir,

When I using HiCzin, I got an error on the step 'Normalizing raw contacts by HiCzin'. The log file and current output files can be found here. Thanks!

valid_contact.csv.txt

hicbin.log

contig_info.csv.txt

[dyxstat]

Hi Jianchang,

Thanks for reporting this problem. This problem comes from your coverage file since the coverages of 1,502 contigs are zero. Therefore, in the normalization, the biases factor \log(coverage_i * coverage_j) becomes negative infinity. For more details about the normalization, please refer to our RECOMB HiCzin paper.

I have updated the hicbin.py and Hiczin.R to solve this problem. Please just replace the original two scripts with the updated version.

Moreover, in the command line, '--thres' denotes the acceptable fraction of incorrectly identified valid contacts, which ranges from 0 to 1. We suggest not selecting a large value as this will discard a lot of 'good' Hi-C contacts. From our experiments, 5% could already help us discard most spurious contacts, which is also our default value.

Best, Yuxuan

[Taojianchang]

Hi Yuxuan,

Thanks for solving the problem and the suggestion on '--thres' setting. I tried again, however, there was an other error occurred. The error log is: "Traceback (most recent call last): File "./hicbin.py", line 185, in ifelse(args.min_binsize, runtime_defaults['min_binsize'])) File "/home/jianchang/software/HiCBin/Cluster.py", line 55, in init self.norm()
File "/home/jianchang/software/HiCBin/Cluster.py", line 89, in norm c = (log(c1*c2)-self.norm_result[9])/self.norm_result[10] ValueError: math domain error".

Best, Jianchang

[dyxstat]

Thanks, Jianchang. This problem is again because the abundances of some of your contigs are zero.

Please update the script 'Cluster.py' to solve this problem.

Best, Yuxuan

[Taojianchang]

Thank you. I got the resulting bins now.