Problem with heritability estimation

[YinLiLin]

Dear Hui Li,

Thank you very much for developing the advanced algorithm and the powerful software tool ‘HEELS’, in which we are pretty interested and want to make a comparison with our proposed algorithm in terms of computational accuracy and efficiency. However, we encountered unexpected errors and could not solve it by now after many tries, thus we seek for your professional suggestions and guidance on the usage of HEELS.

We are trying to evaluate the performance of HEELS on UKB dataset (~320,000 unrelated individuals), and mainly focus on the local heritability estimation of one chromosome (chr2). We first constructed block-wise LD matrices for chromosome 2 using randomly selected 50,000 individuals from UKB, with the block information estimated by Berisa and Pickrell following the descriptions of your published paper, then employ these LD matrices to run HEELS on the summary data of different traits of the remaining individuals (~270,000) via cross-validation. Unfortunately, we always encounter the similar error during the iteration process (i.e., “leading minor not positive definite”), no matter what traits we used in analysis. For convenience, we have attached the running scripts and the LOG file here, please kindly have a check.

We would greatly appreciate any insights, guidance, or suggestions you might have to address this challenge. Thank you again for your time and dedication to creating such a remarkable tool. We shall look forward to your response and seek for a further cooperation.
HT_intscl_fold_1_chr2_heels_block.log

Best,
Lilin