Don Gilbert's recipe for RNA-Seq assembly to gene models
2009 Sept,  gilbertd at indiana edu
=====================================================================================    
Evaluation:

* Using DrosMel reference genes, FlyBase gene version 5.10, and comparison to
PASA EST assembly of 570,000 ESTs (dbEST, March 2008).

* Ref. genes are filtered to remove genes with less than 40 bases of read coverage (>4reads/base)
to evaluate gene models where coverage exists.

* Models are scored for sensitivity and specificity at exon, transcript and gene levels,
using bases matched and bases missed (over and under) of reference genes.  THese are compared
with EST assembly scores.

* Model quality errors are also tabulated by macovsup.

=====================================================================================    
DrosMel all chroms:

All EST asm
#collect_overlaps=108684
ngene=16412; genehit=10400; gperfect=5550; nexon=52759; exonhit=41735
Exon: Sn=69.7; Sp=82.7; Sp2=37.3; Ave=76.2; hit=18053972; miss=3751482; exonbase=25866420; allover=48395156
BestTr: Sn=57; Sp=83.1; Sp2=48; Ave=70.05; hit=23237010; miss=4715178; allbase=40728485;
Gene: Sn=54.5; Sp=85.2; Sp2=25.5; Ave=69.85; hit=12378575; miss=2146236; genebase=22680550;
#overlaps found=41735


RNASeq.fine.genes; macovsup v1.19 
   exons:67608,  introns:31835,  genes:35122,  alt-tr:6839  (excluding poor*)

#collect_overlaps=139547
ngene=16412; genehit=12409; gperfect=4156; nexon=52759; exonhit=40119
Exon: Sn=74.1; Sp=55.2; Sp2=33.8; Ave=64.65; hit=19185234; miss=15521787; exonbase=25866420; allover=56641135
BestTr: Sn=69.5; Sp=71.5; Sp2=50; Ave=70.5; hit=28330955; miss=11288767; allbase=40728485;
Gene: Sn=68.8; Sp=72.9; Sp2=27.5; Ave=70.85; hit=15615566; miss=5776000; genebase=22680550;
#overlaps found=40119

Quality error counts on all RNASeq.fine models
21971 cdspoor     *
 6019 exonabut
12178 exonshort
 1389 intronabut
 3048 intronend
 2316 intronxgap
  991 mixedstrand
 8286 noscore     *
 4223 overpoor    *
   * = classes excluded from tally above

=====================================================================================    

Read and alignment counts per grouping, as count(per-reads)

all: nr=338896759, ni=39, avelen=1552(39), farmate=1836849(0.00542), mated=134908497(0.398), mismate=119680942(0.353), na=413953537(1.22), poor=39471276(0.116), stranded=4618360(0.0136), 

# Cell line (pe = paired end reads, sr = single reads)
dmCMEpe: nr=50095520, ni=5, avelen=185(37), farmate=284143(0.00567), mated=19557742(0.39), mismate=34091560(0.681), na=56144412(1.12), poor=7408476(0.148), stranded=580675(0.0116), 
dmKc167pe: nr=49099950, ni=5, avelen=185(37), farmate=344505(0.00702), mated=17906104(0.365), mismate=23757345(0.484), na=44607619(0.909), poor=7100159(0.145), stranded=577815(0.0118), 
dmKc167sr: nr=46800876, ni=7, avelen=252(36), farmate=0(0), mated=0(0), mismate=0(0), na=79415242(1.7), poor=2621574(0.056), stranded=552195(0.0118), 
dmMLpe: nr=46974396, ni=6, avelen=222(37), farmate=328651(0.007), mated=27902922(0.594), mismate=38088029(0.811), na=68272911(1.45), poor=5734540(0.122), stranded=669985(0.0143), 
dmS2pe: nr=103022042, ni=9, avelen=333(37), farmate=879550(0.00854), mated=69541729(0.675), mismate=23744008(0.23), na=98524849(0.956), poor=7910580(0.0768), stranded=1643909(0.016), 
dmS2sr: nr=42903975, ni=7, avelen=375(53), farmate=0(0), mated=0(0), mismate=0(0), na=66988504(1.56), poor=8695947(0.203), stranded=593781(0.0138), 

# Read length
rl36: nr=46800876, ni=7, avelen=252(36), farmate=0(0), mated=0(0), mismate=0(0), na=79415242(1.7), poor=2621574(0.056), stranded=552195(0.0118), 
rl37: nr=249191908, ni=25, avelen=925(37), farmate=1836849(0.00737), mated=134908497(0.541), mismate=119680942(0.48), na=267549791(1.07), poor=28153755(0.113), stranded=3472384(0.0139), 
rl45: nr=27808540, ni=5, avelen=225(45), farmate=0(0), mated=0(0), mismate=0(0), na=43688500(1.57), poor=8695947(0.313), stranded=98549(0.00354), 
rl75: nr=15095435, ni=2, avelen=150(75), farmate=0(0), mated=0(0), mismate=0(0), na=23300004(1.54), poor=0(0), stranded=495232(0.0328), 

# Read type (PE = paired reads, SR= single reads)
typePE: nr=249191908, ni=25, avelen=925(37), farmate=1836849(0.00737), mated=134908497(0.541), mismate=119680942(0.48), na=267549791(1.07), poor=28153755(0.113), stranded=3472384(0.0139), 
typeSR: nr=89704851, ni=14, avelen=627(44), farmate=0(0), mated=0(0), mismate=0(0), na=146403746(1.63), poor=11317521(0.126), stranded=1145976(0.0128), 

# Run Date
080715: nr=9649105, ni=2, avelen=90(45), farmate=0(0), mated=0(0), mismate=0(0), na=13080945(1.36), poor=3348870(0.347), stranded=31924(0.00331), 
080815: nr=18159435, ni=3, avelen=135(45), farmate=0(0), mated=0(0), mismate=0(0), na=30607555(1.69), poor=5347077(0.294), stranded=66625(0.00367), 
081114: nr=10654850, ni=1, avelen=37(37), farmate=91445(0.00858), mated=8443925(0.792), mismate=1839736(0.173), na=10675788(1), poor=533863(0.0501), stranded=178834(0.0168), 
081121: nr=88856312, ni=8, avelen=296(37), farmate=660804(0.00744), mated=46155339(0.519), mismate=41129024(0.463), na=92058505(1.04), poor=10640414(0.12), stranded=1152538(0.013), 
081202: nr=47130556, ni=7, avelen=259(37), farmate=488576(0.0104), mated=24616410(0.522), mismate=26682125(0.566), na=53577983(1.14), poor=5203868(0.11), stranded=693305(0.0147), 
081216: nr=67593194, ni=6, avelen=222(37), farmate=457444(0.00677), mated=34408898(0.509), mismate=39903606(0.59), na=77922250(1.15), poor=8750030(0.129), stranded=920950(0.0136), 
081223: nr=34956996, ni=3, avelen=111(37), farmate=138580(0.00396), mated=21283925(0.609), mismate=10126451(0.29), na=33315265(0.953), poor=3025580(0.0866), stranded=526757(0.0151), 
090311: nr=4988862, ni=1, avelen=75(75), farmate=0(0), mated=0(0), mismate=0(0), na=8513382(1.71), poor=0(0), stranded=188153(0.0377), 
090512: nr=10106573, ni=1, avelen=75(75), farmate=0(0), mated=0(0), mismate=0(0), na=14786622(1.46), poor=0(0), stranded=307079(0.0304), 

#...............
Value = sum (per-reads)
Keys: 
nr = number of reads/group (2 paired reads are counted as 2)
ni = number of replicate runs/group
na = number of accepted_hits alignments (using tophat/bowtie -k 40 max alignments/read);
avelen = average read length
mated = alignments with valid pairs, 
mismate = alignments with invalid mate, 
poor = under 97% identity (2+ mismatches for 37bp), 
stranded = spliced alignments,
farmate = mates excluded by distance (25Kb in this run, should be increased greatly).

Note: na/nr ratio is 1.0 for PE, but 1.6 for SR, due to way Tophat chucks out 
mate-pair duplicates with longer inner spans.

=====================================================================================