issue while running SMulTiXcan.py. AttributeError: 'Series' object has no attribute 'zscore'

[charliexbrown]

I get AttributeError while running SMultiXcan from the results from SPrediXcan:

python ~/SMulTiXcan.py --models_folder ~/mashr_mod/eqtl/mashr/ \ --models_name_pattern '(.).db' \ --model_db_snp_key varID \ --gwas_file wt.tpmt_gwas4predix.txt \ --snp_covariance ~/mashr_mod/gtex_v8_expression_mashr_snp_smultixcan_covariance.txt.gz \ --snp_column ID_b38 --keep_non_rsid --effect_allele_column ALT --non_effect_allele_column REF \ --beta_column BETA --se_column SE --pvalue_column P \ --metaxcan_folder spredixcan_output/ --metaxcan_file_name_parse_pattern '(.)_(.*).csv' \ --output spredixcan_output/wt.tpmt_multix.csv \ --throw \ --cutoff_condition_number 30 \ --verbosity 4

... INFO - Creating MetaXcan results manager Traceback (most recent call last): File "/home/cbidoli/MetaXcan/software/SMulTiXcan.py", line 93, in run(args) File "/home/cbidoli/MetaXcan/software/SMulTiXcan.py", line 22, in run context = CrossModelUtilities.context_from_args(args) File "/research/rgs01/home/clusterHome/cbidoli/MetaXcan/software/metax/cross_model/Utilities.py", line 162, in context_from_args metaxcan_manager = MetaXcanResultsManager.build_manager(args.metaxcan_folder, filters=args.metaxcan_filter, file_name_pattern=args.metaxcan_file_name_parse_pattern) File "/research/rgs01/home/clusterHome/cbidoli/MetaXcan/software/metax/metaxcan/MetaXcanResultsManager.py", line 35, in build_manager manager = MetaXcanResultsManager(results) File "/research/rgs01/home/clusterHome/cbidoli/MetaXcan/software/metax/metaxcan/MetaXcanResultsManager.py", line 14, in init data, genes, models = _build_data(data) File "/research/rgs01/home/clusterHome/cbidoli/MetaXcan/software/metax/metaxcan/MetaXcanResultsManager.py", line 71, in _build_data if not numpy.isfinite(row.zscore): File "/hpcf/authorized_apps/rhel7_apps/anaconda3-202210/install/lib/python3.9/site-packages/pandas/core/generic.py", line 5575, in getattr return object.getattribute(self, name) AttributeError: 'Series' object has no attribute 'zscore'

I thought it was a conda packages issues but both numpy and pandas are updated (numpy=1.24.2, pandas=2.0.0) so I don't know how to fix this error. Thanks!

[Fnyasimi]

Hi @charliexbrown looks like the zscore column is missing. I suggest you harmonize your gwas and provide the zscore instead of beta and se.

[charliexbrown]

Thank you @Fnyasimi! I figured out the error, in this case I had some extra files in the metaxcan folder that were causing some background noise while performing the analysis. In the end it worked without harmonizing the gwas, but is zscore imputation always recommended in order to perform multixcan analysis? Thank you!