-
Notifications
You must be signed in to change notification settings - Fork 57
/
Copy pathCHANGES.txt
48 lines (24 loc) · 1.99 KB
/
CHANGES.txt
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
===================== Brief and incoplete release notes =====================
Nov 20th, 2019, v 1.0.11
- Implemented Joel Mefford's sparsified BLUP prediction (Mefford, thesis 2018), which we dub LDpred-fast. LDpred-fast may be
suitable for polygenic diseases/traits when LDpred-gibbs fails to converge or is too slow.
- Fixed a negative correlation bug, and now the user must define on what scale the reported effects are.
- Other minor fixes and updates.
Oct 21th, 2019: v 1.0.10
- LDpred-gibbs now reports LDpred-inf effects for SNPs in long-range LD regions (Price et al., AJHG 2008). This
improves convergence of the algorithm substantially when applied to large datasets.
- Added --summary-file flag in score, so that prediction results for all methods could be written out into a file
- Updated integration tests, so that they can be included via pip.
Oct 17th, 2019: v 1.0.8
- Changed package structure to fix pip issue.
- Improved testing framework, and refactoring.
- Enabled testing when installed via pip, using ldpred-unittest commands
- Fixed a bug in LDpred that could improve convergence for gibbs.
- Other minor improvments
Oct 11th, 2019: v 1.0.7
- Improved handling of variants with p-values rounded down to 0.
- Added options to enable effect estimate inference from standard errors (--z-from-se) in coordination step.
- Fixed a serious bug that caused sample sizes in summary stats file not always be used correctly when provided.
- LDpred gibbs can now handle differing sample sizes per variant effects, if they are parsed in summary stats. This can improves convergence and accuracy subtantially, but slows the gibbs sampler (approx 5x). It can be avoided by setting --N flag.
- If the heritability is not provided when running LDpred gibbs and inf it now estimates it separately for each chromosome, which can improve accuracy if summary stats sample size is large. To avoid this behavior, either set --h2 flag, or use --use-gw-h2 flag.
- Improved reporting and other things.